Read more

I am long MRLN equity. As of the October 23 close, MRLN trades at $0.88. This post is research synthesis, not investment advice.

There is a word sitting inside Merlin's website right now that Merlin has never said out loud.

That word is redtail. It ships to every browser that loads merlinlabs.com, in a list of page slots, wedged between Condor and the main Merlin Pilot page. Nothing lives at the address yet. On October 10, about four days before that slot first appeared, Merlin flew a 1976 Beechcraft King Air C90 autonomously for the first time and called it the first flight of the next generation of Merlin Pilot.

Thesis in two sentences. The King Air program is real flight-test progress on the software, and the most likely job of the unlaunched Redtail page is to sell that software as a tactical, ISR and maritime product line. What the record does not show yet is a customer, a dollar or a Navy program behind it. A stock trading at $0.88 is pricing the financing calendar, not the flight-test calendar.

A reader on X, @1MoreSmithHere, recently put three good questions on the table about all of this. They deserve a real answer with receipts, so this runs long. Every claim below is sorted into verified fact or hypothesis. Anything I could not verify is flagged as such.

I. How I Found a Page Merlin Has Not Published

I read Merlin's website the way I read its filings. Not the words on the page. The plumbing under it.

The site is a static build fed by a content system. Every rebuild ships a bundle of JavaScript to your browser, and that bundle carries a source map naming the files it came from. On October 16, pulling a fresh copy of the bundle for a build-by-build diff, I found a map of what the code calls replacement pages. Its order: homepage, condor, redtail, pilot, then the corporate pages, then the legal pages, then resources. Redtail also sits in a reserved list of page names that no ordinary page is allowed to take. A code comment above the map reads "Editable section contracts for rebuilding the existing site." You can see the string in Merlin's production JavaScript chunk, and it is still there as of this morning.

Three things about that slot matter.

First, it is product tier. Condor's live page is titled Condor, with the subtitle "Merlin Pilot for Large, Multi-Crew Aircraft," and it uses a hero size no other page on the site uses. Condor went live the same day as the May 14 8-K. Redtail sits next to it, ahead of the generic Pilot page. A program or partner announcement would be a news post. This is a product slot.

Second, it has never been live. The Redtail page data returned a 404 on October 16 and returned a 404 again today. Its sitemap grew from 91 to 94 entries since then, with no Redtail entry. On the Wayback Machine there are zero captures of any Redtail path on any Merlin domain, ever. Nine archived builds of the site back to March 17 do not contain the word. It first ships in a build dated between October 10 and October 15, anchored to October 14.

Third, it can go live without an engineer. Only a content record that claims the redtail slot creates the page. A marketing editor can publish it from the content system and the site rebuilds itself. No new code, no new chunk. If you are watching for a launch, watch the page data, not the JavaScript.

Then I ran the negatives, because a word in a code file is a lead, not a finding.

• USPTO trademark search: 139 RED TAIL and REDTAIL family records through filings dated October 15. None is owned by Merlin. None was filed by Merlin's trademark counsel. That is weak evidence, since Merlin never filed for CONDOR or MERLIN PILOT either.

• EDGAR: zero hits for Redtail in any Merlin filing. Here the control matters. Condor never appeared in an SEC filing before its launch day either.

• FAA registry: 43 aircraft with Redtail or Red Tail owner names. None is Merlin. None is a King Air, an L-39 or a Caravan.

• Job board: 24 live postings read in full. Zero hits.

• LinkedIn: 317 profiles listing Merlin, plus the company page. Zero hits.

Sum it up. Merlin reserved a product page named after a hawk, days after the King Air flight and one day before it revealed an L-39 jet, and it has told nobody what goes on that page.

One naming note. On X it has been called Project Redtail. In the code it is just redtail. Merlin does use project code names for corporate work, and its de-SPAC deck carried the file name Project Midnight. Nothing public attaches the word project to Redtail, so I use plain Redtail below.

II. What Merlin Actually Said on October 12

Strip the commentary away and start with the primary text.

On Monday, October 12, @MerlinAero posted that over the weekend Merlin completed the first flight of the next generation of the Merlin Pilot, and that, to its knowledge, it was the first autonomous flight of a King Air C90, which it called "a workhorse of the global ISR fleet." According to the post, the new software is designed to use sensors "to make mission choices typically reserved for humans," aimed at maritime search and patrol. Saturday's flight, it said, kicks off "an intense flight test campaign." This version "lays the groundwork for what we believe could be the United States' first fully autonomous (versus pre-programmed or remotely piloted) operational group 5 platform next year, with thousands of pounds of payload and thousands of miles of range." Its last line promised a foundation for "other tactical applications" on platforms "markedly different from what we've previously flown."

A forward-looking statement reply followed on the same thread and named "a potential operational Group 5 platform next year" as one of the covered statements. That matters. Lawyers read that post before it went out.

Matt George added context the same morning. In one post he argued that aircraft like the MQ-9 and Global Hawk "often aren't very autonomous" because they fly pre-planned missions or need a team on the ground. In another he called the flight a step toward "the first large aircraft that can truly fly autonomously for useful missions."

Now the receipts under those words.

Airframe. On the FAA registry entry for N122K: a Beech C90, serial LJ-707, built in 1976, owned by Merlin Labs Inc., registration issued May 14, 2026. Its engine field reads Pratt & Whitney Canada PT6A-135A, a newer engine than the one the airframe was born with. Its airworthiness block reads Experimental, Market Survey, Research and Development, dated August 11, 2026.

That August date is the most important fact in this section. An aircraft flying a removable test kit can stay on its standard certificate. An aircraft carrying a permanent flight-control modification has to leave the standard category, and an experimental research certificate is the usual path. N122K changed certificates on August 11. So the King Air crossed from baseline data collection into a modified test article in August, two months before the announcement.

Flying. I pull ADS-B for every Merlin tail. N122K flew three sorties at Centennial, Colorado on May 24, went dark for 96 days, flew a maneuver box in the Shenandoah Valley on August 28, ferried to Quonset on August 29 and flew from Quonset on September 2. In September the team moved to North Central State Airport in Smithfield, Rhode Island, which Merlin announced as its new US flight base on October 20. Between September 29 and October 14 the King Air flew six offshore sorties, about two a week, each a little longer and a little higher than the last. October 10 was a Saturday, outside the team's usual weekday rhythm. The announcement came Monday morning.

Sequence. On October 13 and 14 the company posted dated teasers reading "October 15, 2026." On October 15 it revealed an Aero L-39C Albatros as its first jet testbed. Its L-39 announcement page lists exactly three "active Merlin Pilot integration programs": the Caravan, the King Air C90 and the C-130J. King Air role, in Merlin's words: "Long-endurance ISR, maritime missions, and contested logistics across extended ranges." KC-135 is not on that list.

What is missing. No 8-K. No wire release. The Q2 10-Q, filed August 14, three days after the King Air's certificate changed, does not contain the words King Air. Meanwhile the homepage has carried a "Maritime reconnaissance" mission domain since the March rebuild, reading "Merlin is testing the King Air for Maritime ISR," per the homepage page data. Management had the King Air in the plan for months. It never made the filings.

That gap tells you how management classifies the King Air today. A test program and a marketing story. Not yet a disclosed customer program.

III. Why a 1976 King Air

People keep asking why a C90 and not something newer. Economics and the cockpit answer it.

Cheapest serious twin you can fly. The King Air 90 family is the long-running small end of the King Air line. Even its newest version, the C90GTx, has a useful load of about 3,280 pounds on two PT6A-135A engines at 550 shaft horsepower each. Older C90s trade in volume on the used market, and lessors keep dozens of them. A 1976 airframe re-engined with -135As is about as low a capital cost as a pressurized twin turboprop gets. Since the de-SPAC, Merlin's capex has gone to used test airframes, not production tooling. This fits.

A twin, finally. Merlin's original flying was single-engine Caravans. Two engines force the autonomy stack to handle asymmetric thrust, engine-out decisions, engine-out climb performance and twin approach logic. Every large transport Merlin wants to sell into is multi-engine. A C-130J has four. A twin is the cheapest place to build and test that decision logic.

An old cockpit on purpose. A 1976 King Air has a mechanical control system and analog-era instruments unless someone has gutted the panel. Per Merlin's own technical disclosures, its architecture back-drives the legacy controls with electromechanical servos and reads the panel with cameras. An old airplane is the hardest test of that design. Old is also representative of the fleets Merlin wants to retrofit, because most of the world's military utility and ISR turboprops are not new.

Everywhere in government service. "Workhorse of the global ISR fleet" is marketing, but the family tree is real. Navy pilots know the King Air 90 as the T-44. Its larger cousins, the Super King Air 200 and 350, became the C-12 for the Army and Air Force, the UC-12 for the Navy and Marines, and the base for a long list of special-mission ISR conversions worldwide. Australia's maritime surveillance contractor flies King Air 200s today. Pick a King Air and you are building on the most common special-mission turboprop on earth.

A modest certification step. N122K flies on an experimental research certificate. That lets Merlin modify and test the airplane without a type certificate change. The market survey category also lets it fly demonstrations for prospective customers. Selling the system is a different matter. On September 9 Merlin withdrew its small-aircraft civil certification application in New Zealand, so the small-aircraft civil route is no longer a near-term catalyst. A King Air product would live first in the military world, where the customer runs its own airworthiness process.

Now the Group 5 math, where marketing gets ahead of the airframe. In Defense Department terms a Group 5 uncrewed aircraft is anything above 1,320 pounds that operates above 18,000 feet. A King Air clears that easily. "Thousands of pounds of payload" fits too, once you pull the crew seats and fill the cabin. "Thousands of miles of range" is the stretch. For the C90GTx, published maximum range is about 1,260 nautical miles, roughly 1,450 statute miles, and a 1976 airframe will not beat a new one. Hitting thousands of miles and thousands of pounds on one flight would take auxiliary fuel trading against payload, a bigger King Air, or a different airplane. The post's own last line points at "platforms that are markedly different." My read: the Group 5 claim is about the software, and the operational airframe for it is not necessarily the airplane flying today.

IV. Who Else Is Teaching King Airs to Fly Themselves

Read the claim carefully. First autonomous flight of a King Air C90, to its knowledge. Not the first autonomy on a King Air.

Garmin got there first in a different sense. Its Autoland and Autothrottle have been FAA certified for retrofit on G1000 NXi-equipped King Air 350s since August 27, 2025. Autoland is an emergency function. If the pilot is incapacitated, it picks an airport, flies there and lands. Autoland flies no mission and makes no sensor-driven patrol decision. Nor does it replace a crew on purpose. Still, the King Air community already accepts a certified box that lands the airplane alone. Garmin owns that cockpit on the newer airframes.

Honeywell runs an experimental testbed fleet that includes a King Air B200, alongside a PC-12 and a regional jet, so the avionics majors are testing in Merlin's classes with their own airplanes. Reliable Robotics flies remotely piloted Caravans with Air Force money. In the special operations world, Beacon AI holds a SOCOM prototype agreement for crew-workload autonomy, signed April 15, 2026, in the same budget neighborhood as Merlin's C-130J work.

Against that field, the differentiator is specific. Onboard autonomy with no datalink dependency, aimed at flying a mission rather than landing in an emergency, on aircraft that already exist. That is a real niche. Every rival named above already has a certified product or a funded contract. Merlin's King Air has neither yet.

V. A Navy Trainer Hiding in the Type Certificate

Here the thread gets interesting, so start with history.

Officially, the T-44 Pegasus is the Navy's military King Air 90. It flew its first operational flight in April 1977, and 61 were delivered between 1977 and 1980. Each original T-44A ran two PT6A-34B engines at 550 shaft horsepower. After 2006 the Navy upgraded the whole fleet's avionics into the T-44C configuration. For almost five decades it has been where Navy, Marine Corps and Coast Guard multi-engine pilots learn to fly maritime aircraft, mostly at Naval Air Station Corpus Christi under Training Air Wing Four.

Its replacement is arriving now. In February 2023 NAVAIR picked Textron Aviation for a multi-engine training system built on the King Air 260, designated T-54A, with up to 64 aircraft on the contract. Janes reported a first lot of ten aircraft worth about $113M, and reported that T-44 retirements begin six months after the first T-54 delivery. First T-54As arrived at Corpus Christi in April 2024, and Naval Aviation News described them as replacing the aging T-44C. NAVAIR said in 2025 that deliveries run through calendar 2026. In December 2025 the first four student aviators completed T-54A training.

So the verified version of the T-44 pattern is simple. Merlin is flying the civil version of a Navy trainer. On a schedule that overlaps with Merlin's King Air campaign, the Navy is retiring that trainer. Textron, which builds the replacement, is already a Merlin counterparty through the SkyCourier UX work. All three statements are true. None of them, alone, connects Merlin to the Navy.

VI. The Thread, and Where It Goes Further Than the Record

Credit where due. @1MoreSmithHere has been one of the sharper retail readers of the Merlin record all year. Early in October he posted SOCOM budget justification material on AC/MC-130J mission systems, and after the King Air post he laid out the breadcrumbs in Merlin's own wording. In a recent thread he argued that Redtail refers to the Navy's T-44, the C90 variant now giving way to the T-54, and asked three questions. In my paraphrase:

1. Does the C90 choice signal a new or expanded Navy opportunity, since the T-44 is a Navy platform and the Air Force's King Air-family aircraft is the larger C-12?

2. Does the King Air let Merlin keep moving on the C-130J program if access to a C-130J airframe for near-term flight testing is constrained?

3. Could retiring T-44s be repurposed for ISR or other missions that benefit from Merlin Pilot?

Those are his hypotheses, not Merlin's claims, and he framed them as questions. I treat them that way. For each one: what the public record supports, what cuts against it, where I come out.

One structural point first. Any T-44 link is an inference from the airframe type. Nothing Merlin has published mentions the T-44, the T-54, Corpus Christi or the Navy training command. Its King Air post talks about ISR and maritime patrol, which are operational missions, not training. That does not kill the thread. It sets the burden of proof.

VII. Question One: Is This a Navy Door?

What supports it.

• Navy-shaped mission language. Maritime search and patrol is the headline mission in the October 12 post. On the homepage the King Air domain is "Maritime reconnaissance." On the L-39 page it is "maritime missions." Two October blog posts on Palau, including a visit from President Whipps, frame autonomy as a way to patrol an exclusive economic zone the size of Texas. A sea services and partner-nation problem.

• Fresh money with that exact label and no name on it. In the House report on the FY27 defense appropriations bill, H.R. 9495, H. Rept. 119-715, appropriators add $35M to a Navy research line for "autonomous maritime patrol aircraft." I found no larger unclaimed airframe-autonomy add in the House marks. It names no performer.

• A Navy chair on the board. Merlin's seven-person board, named April 16, includes Kenneth Braithwaite, the 77th Secretary of the Navy.

• A maritime seat on the org chart. Merlin organized general manager roles by customer axis. Its tactical seat has had a leader since August, a hire from AeroVironment. Its maritime seat has been posted since December 2025 and is still open. Recent arrivals also include a former Navy test pilot and a former Coast Guard aircraft coordinator.

• A CEO who keeps saying there is more. On October 7 he wrote that it is "insanely frustrating" to have things the company cannot talk about yet.

What cuts against it.

• T-44 means training. The Navy's need in that airframe class is pilot training, already being met by the T-54A, a new aircraft with a modern glass cockpit. No service retires a trainer and then buys autonomy for the retiree's training mission.

• No Navy dollar. USAspending shows no Navy contract to Merlin, ever. Its entire federal contract ledger is the SOCOM IDIQ, a small FAA research award, two Air Force SBIR awards and a $500 Missile Defense Agency order. Since October 1, 2025, new federal money to Merlin totals that $500.

• House-only and unattributed. No lobbying tie connects Merlin to the Navy add; its known appropriations ask is its own SOCOM line. Government funding runs on a continuing resolution through December 11, and a full-year resolution would erase every add in the House bill, Merlin's own included.

• An empty Navy page. Before 2026, Merlin's site carried six customer pages, including one for the Navy. Every one was a template with no content.

• A weaker Air Force contrast than it looks. On airframe sizes the thread is right: the Navy trained on the C90-family T-44 and the Air Force flies the bigger C-12. But the Navy also flies the bigger King Airs as the UC-12. Choosing a 90 over a 200 or 350 reads to me as cost and availability, not service selection.

Where I come out. The maritime angle is real, and the Navy is the natural customer for it. The T-44 is not what makes it real. If Merlin has a Navy door, it is the autonomous maritime patrol add, partner-nation EEZ patrol and the maritime GM seat, with the King Air as the demonstration airframe for all three. Proof would be a performer named on that Navy line, a GM Maritime hire, or a Navy contract action under Merlin's name. Until one of those prints, I score a new Navy program as plausible and unproven.

VIII. Question Two: Can a King Air Stand In for a Hercules?

Best of the three questions, because it goes at the program that carries the stock.

Where the C-130J program stands, verified.

• Vehicle. IDIQ H9240824D0001 is SOCOM's "Automated Crew & Workload" contract, a $105M ceiling running from June 2024 to April 2029. The May prospectus says it covers the initial adaptation and can be extended to low-rate and full-rate production for all AFSOC fixed-wing aircraft.

• Money. Two task orders have ever been issued against it. A kickoff order for $7,535, and the real one, H9240824F0010, for "PDR & CDR," with $15,955,778.32 obligated and a period of performance that ran out on August 31, 2026. Defense contract reporting lags, so a new order could exist and not show yet. On the public record, there is no third order.

• Reviews. Merlin completed the preliminary design review on March 5 and the critical design review on June 4. According to the CDR release, the program moves into aircraft integration, then ground testing, then flight demonstration.

• Airframe problem. When the PDR closed, Merlin's CTO said the AC-130J the company planned to fly first was no longer available. He also told Aviation Week that the C-130J's existing autopilot servos lack the authority Merlin needs, which puts a flight-control hardware change inside the job. After the CDR, AIN reported that no specific C-130J had been designated for integration, with a company spokeswoman citing SOCOM's operational demands. Those demands are concrete. Per the Congressional Research Service, two MC-130Js were destroyed on the ground in Iran in April during the rescue of a downed F-15E crew, and AFSOC had already accepted its final MC-130J off the production line. A small fleet that just lost two aircraft in combat does not hand a test article to a software vendor quickly.

• Promise. Merlin's Q2 10-Q still reads "We currently anticipate initial customer deliveries within the next year" under the C-130J contract. That sentence has rolled forward from filing to filing. On my ledger of Merlin's dated public promises, the median delivered item has taken about 1.85 times as long as promised.

So the thread's premise holds. Airframe access is constrained, on the company's own account.

What the King Air can carry for the C-130J.

Matt George answered part of this himself on October 15: Merlin is building a pilot that is "aircraft and mission agnostic," "the core of everything we're flying today, and the foundation of our work on the C-130J program." Merlin's one brain, any aircraft post makes the same argument at length. Take that at face value and here is what plausibly transfers:

• Autonomy core. Decision logic, flight management, contingency handling, ATC voice and the sensor-driven mission behaviors the King Air post describes. Software matures with flight hours on any airframe.

• Multi-engine logic. Engine failure, asymmetric thrust, performance-limited decisions. A four-engine turboprop needs all of it.

• Test infrastructure and evidence. Procedures, data pipelines, safety cases and the habits of a flight-test organization that flies twice a week. N122K and one Caravan now account for about two thirds of Merlin's US flight hours.

• Hardware-in-the-loop confidence. Merlin's CDR release put hardware-in-the-loop simulation ahead of aircraft integration. Flight data from a real twin feeds better simulation.

What the King Air cannot carry.

• C-130J flight-control integration. On the C-130J, a Honeywell digital autopilot drives Woodward servos. If those servos lack authority, someone has to qualify new hardware on a C-130J. No King Air flight touches that.

• Data rights and the integrator. Lockheed designed the C-130J flight deck. Any retrofit path runs through partners that modify C-130s and through the Air Force airworthiness process. That work needs the real aircraft and the real customer.

• SOCOM's own milestone. SOCOM pays for its program on its airframe. No King Air sortie counts as a C-130J test point, and none will trigger a task order by itself.

Where I come out. Yes, the King Air keeps the software moving while the C-130J sits in the integration queue, and that is the most defensible reading of why Merlin bought it. No, it does not retire the C-130J-specific risks, which are hardware, data rights and airframe access. My internal model for a first C-130J flight in Merlin configuration has its median in early 2028, with close to zero odds in 2026. A King Air does not move that date. A designated airframe would. So would a new task order under H9240824D0001, or an Aviation Week-grade update on the servo fix.

IX. Question Three: A Second Life for Retiring Pegasus Airframes?

Most speculative of the three, with the best primary-source support. Not for the T-44. For the idea.

What Merlin has written down. In the business section of its May 13 prospectus, Merlin lists four tactical autonomy areas. Here is the second, verbatim: "we work on automating older aircraft quickly to provide highly effective, optionally-crewed tactical systems, demonstrating the flexibility of our Merlin Pilot system in tough conditions, through fully automated surveillance and strike capabilities."

An earlier draft of that sentence was more specific. In the draft registration statement Merlin submitted on October 3, 2025, the same sentence ends with a code name, "(the "Fireships" concept)." January's S-4 deleted the code name and kept every other word. Fireships was never aerial firefighting. It was the concept of turning old crewed aircraft into optionally-crewed surveillance and strike systems.

Put the two together and the thread's third question stops sounding wild. On paper, the filings describe a business of converting older aircraft into optionally-crewed ISR platforms. A fleet of retiring King Air 90s is, on paper, exactly the kind of older aircraft that sentence describes. Merlin is flying one. Its October post promised an operational Group 5 platform next year.

What cuts against the T-44 version.

• Airframe age. T-44s were delivered between 1977 and 1980 and spent their lives in the training pattern. Trainers age in cycles, landings and pressurizations, not just hours. A retiring T-44 is cheap. It is not young.

• Disposition path. I found nothing public on where retiring T-44s go: storage, sale, transfer to another agency or a foreign military sale. Without that, any T-44 repurposing case rests on a supply chain nobody has described.

• Customer. An optionally-crewed ISR aircraft needs an operator and an airworthiness authority. No service has published a requirement for an optionally-crewed King Air. Real demand signals today are the Navy maritime patrol add, partner-nation EEZ patrol like Palau, and allied maritime surveillance contracts where King Air 200s already fly with crews.

• Fireships track record. Both federal contracts ever named FIRESHIPS, small Air Force research awards, went to Reliable Robotics, Merlin's nearest Caravan rival. Both closed by mid-2024 with no follow-on. Merlin has never held one.

• Strike money. Attritable and strike lanes in the FY27 budget belong to hardware builders. A software company without an airframe sub-attaches. It does not prime.

Where I come out. The concept is Merlin's, in writing. The King Air is the right test bed for it. The T-44 specifically is a clever fit on airframe type and an unsupported fit on everything else. If an optionally-crewed King Air becomes a product, I expect the first customer to be a partner nation or a special-mission operator buying from the large pool of civil King Airs, not the Navy reusing its trainers. A named foreign customer would change that. So would a Navy requirement for optionally-crewed patrol, or a T-44 disposition notice.

X. My Read on Redtail

Here is how I weight the readings today. Judgment calls from the public record, not numbers anyone at Merlin gave me.

A tactical product family: a little better than even odds. "Merlin Pilot for tactical aircraft," built around the King Air for ISR and maritime patrol, with the L-39 as the jet branch and Group 5 as the end state. Evidence: a product-tier slot beside Condor; a homepage maritime domain with no product page; a Pilot page roadmap that lists Drone as a future tier on the Merlin Pilot page; timing between the King Air flight and the L-39 reveal; a tactical autonomy GM hired in August; and L-39 copy that already uses the "Merlin Pilot for" construction Condor uses. Then the bird ladder. A merlin is a small falcon. A red-tailed hawk is the mid-size soaring patrol raptor. A condor is the largest bird in the sky. Merlin names products like a birder.

An L-39 or jet-teaming product: about one in seven. Matt George's October 15 thread talks about collaborative combat aircraft and smarter munitions, and the Red Tails heritage is a fighter heritage. Against it: the L-39 got a news post rather than the product slot. The jet also still sits on an exhibition certificate.

A program or customer page: about one in seven. Defense programs love heritage names. SOCOM itself funded a 2011 research effort called REDTAIL, awarded to a contractor now owned by CACI. Against it: a product-tier slot is the wrong vehicle for a program page, and no 2025 or 2026 program by that name appears in any public contracting record I can find. The thread's T-44 reading belongs in this bucket, at its low end.

A working name that launches as something else: the rest. Merlin does not trademark product names, so a slot can hold a code name. Against it: the URL is public. If it goes live, the name is the brand.

Condor gives the launch template. Its page assets went up May 12 to May 14, and the page went live with the May 14 8-K and the Q1 release. If Redtail follows that pattern, the launch rides the Q3 release, which has to land by November 16. A pattern, not a schedule.

XI. The Part the King Air Does Not Change

Now the uncomfortable section. This month's flight program is the most interesting thing about Merlin. Share price runs on something else.

MRLN closed at $0.8776 on October 23. Its first close below $1 came on October 21, per the daily price history, after starting the month at $1.81. At the 101.07 million basic shares on the Q2 10-Q cover, that is a basic market cap of about $89M. Cash was $183.55M on June 30, with no debt. Quarterly revenue was $2.19M, down from $3.11M a year earlier. Six sell-side targets still published average about $7.92, per the consensus page, after TD Cowen cut to $4 on October 9.

Nobody is confused about the flight program. The market thinks the next financing decides who owns the company, and on the structure as filed, that is a fair worry.

• A participating preferred on top. Merlin's Series A preferred carries a 12% paid-in-kind dividend. My model of its accrued value is about $238.6M at September 30. Because the preferred participates in a sale, common receives nothing until total equity value clears roughly $271M. Terms sit in the certificate of designations filed with the March 20 8-K.

• A ratchet with no floor. People quote $5 as a floor, but it floors only the scheduled conversion price reset. Any cumulative raise above $500K priced below the conversion price re-strikes both the preferred and the Series A warrants to the new price, with no floor. An April 29 8-K papered the financing that set this up, and its issuer standstill lapses around October 29.

• A listing clock. Under Nasdaq Rule 5550(a)(2), thirty consecutive business days below $1 brings a deficiency notice and a 180-day cure period. Counting from October 21, the thirtieth business day falls on December 2.

So here is the math on the King Air line itself, the honest way to value a hawk-named web page.

• Bear case. Redtail launches as a marketing page while N122K stays a test aircraft. The line earns nothing in FY27 beyond what already sits inside the SOCOM vehicle. Zero incremental revenue.

• Base case. The King Air becomes the demonstration aircraft for one funded maritime or partner-nation effort, sized like Merlin's past research awards. A few million dollars of FY27 revenue. That matters to a company that booked $2.19M last quarter, and it does not move the capital structure.

• Bull case. Merlin is named on the Navy maritime patrol add or wins a partner-nation patrol program with a production path, and an optionally-crewed King Air becomes the second funded program next to the C-130J. In that world, Redtail is worth a page.

None of those three cases fixes the preferred. Per share, the King Air bull case adds far more value than the bear case, but every dollar of it sits behind the preferred and the ratchet. Structure first. A product launch helps the company raise money on better terms. The terms stay the terms.

XII. What Would Change My Mind

What would make me take the Navy and T-44 readings seriously:

• Merlin named as performer on the FY27 Navy "autonomous maritime patrol aircraft" line, or any Navy contract action under Merlin's name in USAspending.

• A GM Maritime hire, or that seat closing quietly.

• A Redtail page that says "Merlin Pilot for" anything maritime, ISR or Group 5, or that names a Navy, Coast Guard or partner-nation customer.

• A public T-44 disposition notice that routes airframes to a contractor or a foreign partner.

What would make me mark the King Air down:

• Flight cadence stops after the announcement. Two sorties a week is the baseline. A month of silence would say the October 10 date was set by the communications calendar, not the test card.

• The Redtail slot disappears in a future code deploy. That is the kill signal, checkable on every rebuild.

• Q3's 10-Q again omits the King Air entirely while repeating the C-130J delivery sentence.

My thesis breaks at the capital structure, not the airplane. If Merlin raises money after October 29 in a deal that triggers the ratchet, per-share value for everything discussed here falls with it, and I would say so plainly. If Merlin instead lands a strategic investment from an operating company, which the terms treat as an exempt issuance, the preferred and warrants stay where they are and the King Air story gets the time it needs.

XIII. What to Watch, With Dates

• Any day: the Redtail page data turns from 404 to live. Watch for another pair of dated teasers on the company page first. Merlin used that exact pattern for the L-39.

• Any day: a first L-39 sortie from Smithfield, and the jet's registry entry moving from exhibition to research and development.

• October 27: next FINRA short interest print.

• October 29: issuer standstill under the April purchase agreement lapses. From here, any sub-conversion-price raise above $500K trips the ratchet.

• October 30: Russell rank day. A close below $1 keeps MRLN out of the microcap reconstitution math.

• By November 16: Q3 10-Q. Read it for three things. Cash at September 30 against my model of about $155M. Any King Air or tactical autonomy language. Any subsequent-event financing.

• Mid-November: 13F filings for the September quarter, the first clean look at whether the PIPE buyer sold.

• December 1: next PIK accrual on the preferred.

• December 2: thirtieth consecutive business day below $1 if the price does not recover, and the earliest window for a Nasdaq deficiency notice.

• December 11: continuing resolution expires. A full-year resolution erases the House adds, including the Navy maritime patrol add and Merlin's own SOCOM line.

Bottom Line

@1MoreSmithHere asked the right three questions. My answers, one line each.

Is the King Air a Navy door? Maritime is real and the Navy is its natural buyer, but the T-44 link is the airframe type rather than a program. No Navy dollar exists yet.

Does the King Air keep the C-130J moving? It keeps the software moving, which is the main reason to own one. C-130J hardware, data rights and airframe problems only clear on a C-130J.

Could retiring T-44s become optionally-crewed ISR aircraft? Merlin's prospectus describes that business in plain words. On paper the T-44 is a neat fit. Everywhere else it is unproven.

And Redtail is most likely the page where Merlin sells all of it. The flight program earned the attention. The capital structure still decides the stock.

I am long MRLN equity. This is research synthesis, not investment advice. You should not buy or sell securities based on anything I write. I am not a registered investment advisor. I do not owe you a fiduciary duty. My conclusions could be wrong in ways I have not anticipated. Financial projections are model outputs based on publicly available data. They are not guarantees. Do your own due diligence.

The X thread discussed here is third-party speculation credited to its author. Merlin made no such statement.

More in this series

MRLN: The Complete Thesis

The UAE Deal Changes Everything Merlin Was Supposed To Be

$MRLN: Merlin's Clock After the Pivot: What Moved, What Didn't, What We're

$MRLN: What Management Actually Said

What We Are Waiting For

$MRLN Part 13: 6 Weeks of News

MRLN: Part 4 - The Board, the Preferred Supernova, and the Catalyst Nobody

$MRLN Part 3: The DAWG, the $54.6B, and Why Merlin Is Positioned to Win the

The Discount Went Extreme While The Demand Started Becoming Law

The Voice on the Frequency - My Flight Demo

Alyeska Wrote Both Checks and the Market Missed It

Merlin Intelligence - Live OSINT Tool

Alyeska Is Underwater. None of the Thesis-Breakers Have Hit

$MRLN: I built Merlin World

I Tracked Every Merlin Aircraft for 180 Days. The Testbeds Just Peaked

$MRLN: The Tape Says $3.88. The Calendar Says August.

MRLN's $105M Ceiling Is Not the Revenue, It Is the Authorization

MRLN Part 10: The Deep Update

The Gate Cleared and the Tape Forgot

Disclosure: Shawarma Capital has no position in $HAWK, long or short. This is research, not a recommendation. This is the first part of our HawkEye 360 series.

HawkEye 360 sells something very few companies on earth can sell: the location of a radio transmitter, measured from space. A ship that turns off its transponder still runs its navigation radar. A jammer blinding GPS over the Baltic is still a transmitter. HawkEye flies its satellites in clusters of three, measures the tiny differences in when and at what frequency each satellite hears the same signal, and turns that into a point on a map. The industry name is RF geolocation. The customers are governments, the US intelligence community first among them.

Five months after its IPO, the stock is down 50% from the $26 offer price and sits within a dollar of its all-time low. Almost none of that is about HawkEye. The whole space-data group is down 55% to 70% since the SpaceX listing in June pulled money out of the sector, the IPO lock-up released supply into a falling tape, and investors sold first and read the filings later.

We read the filings. We spent two weeks and several dozen research passes on HawkEye: every federal award record we could find, the Navy's task orders line by line, Japan's defense budget going back to 2023, the company's FCC filings, the orbital data for all 40 of its objects in space, and the NRO Director's written answers to Congress. What came out is a company whose demand is better documented than the market realises, whose catalysts are dated and close, and whose stock is priced as if none of it matters.

I. The demand is on the record

Start with the strongest fact in the whole story, because almost nobody quotes it.

In May 2025, the House Armed Services Committee held its annual hearing on national security space programs. In the written questions for the record, then-NRO Director Chris Scolese answered a question on commercial radio-frequency data. His numbers:

• In May 2024, NRO's commercial RF providers conducted an average of 56 collections a week. In May 2025, they conducted 342.

• In May 2024, those providers delivered an average of 2,840 coarse RF geolocations a week. In May 2025, the figure was 139,140.

• In May 2024, in his words, "the only commercial RF provider delivering data was HawkEye 360."

That is a six-fold increase in tasking and a 49-fold increase in delivered geolocations in twelve months. It is the clearest public evidence anywhere that the US intelligence community has moved commercial RF from experiment to habit, and HawkEye was the company that built the habit.

The same testimony contains a second line that matters even more for the size of the market. The NRO said the data is "procured under the end user's authorities, through NRO's contract vehicles." In plain English: the NRO runs the contract, but the agency that wants the data pays for it. The Government Accountability Office counts the NRO's own commercial-RF budget at an average of $0 to $20 million a year over fiscal 2021 to 2025, with a peak of $55 million. That number has been used to argue the market is small. It is the wrong number. Because every combatant command and agency that wants RF data can fund it through the NRO's vehicle, the addressable spend is the sum of their budgets, not the NRO's.

Then, on August 17, 2026, the NRO announced the Commercial Radio Frequency Capabilities Augmentation contract, CRFCA, and awarded it to HawkEye. The NRO described it as an operational vehicle "designed to scale without contract value caps." It is the production contract that the whole commercial-RF industry has been working toward for four years. HawkEye is the company that got it.

II. The revenue has doubled, and the mix is getting better

HawkEye reported $49.8 million of revenue in the second quarter of 2026, up 87% year on year. First-half revenue was $99.6 million, roughly double the first half of 2025. Guidance for the full year is $215 million to $220 million, and management held it in August.

Two engines are doing the work.

International is the fastest-growing business HawkEye has ever had. International revenue went from $17.1 million in the first half of 2025 to $41.9 million in the first half of 2026, up about 145%. On October 15, HawkEye announced $36 million of new multi-year international contracts signed in the third quarter alone, across more than ten nations in the Indo-Pacific, the Middle East, Europe and the Americas. That followed a roughly $18 million Middle East award in September and a multi-year Indian Ocean contract in July.

ISA adds a second US franchise. HawkEye bought Innovative Signal Analysis, a Richardson, Texas signals-processing contractor, on December 18, 2025. It contributes about $14 million to $15 million a quarter, it brings Space Force and Air Force research relationships HawkEye did not have, and it is one of three awardees on a Space Rapid Capabilities Office program for space-based radar warning.

The one soft spot is the legacy US data business, which slipped about 13% in the first half as the NRO's older study contracts wound down ahead of CRFCA and the government ran on a continuing resolution that froze new starts. That is the line the bears point to. It is also the line that CRFCA is built to refill, which is why it is the single most important number to watch on the November 12 call.

III. CRFCA: the vehicle is built, the money comes next

CRFCA is a single-award, sole-source, firm fixed price production contract with a one-year base and two option years. Its template is the NRO's August 5 award for commercial radar to Capella, ICEYE US and Umbra, which Umbra publicly described as "a firm/fixed price, sole source contract." HawkEye graduated into CRFCA from the NRO's earlier study program. No other RF provider has.

"Without contract value caps" removes the ceiling that capped the earlier study contracts. What fills the vehicle is appropriated money, and that is on a clock: the government is on a continuing resolution through December 11, 2026, and the fiscal 2027 bills are still moving. Two things in those bills are already pointed in HawkEye's direction:

• Section 428 of the House intelligence authorization bill creates a pilot for other agencies to buy directly from the NRO's commercial contracts. That is the end-user funding model from Section I, written into law.

• The Pentagon's push toward autonomous and space-based sensing, from Golden Dome to the new Autonomous Warfare Command, runs on exactly the kind of persistent, passive cueing an RF constellation provides.

We see three paths for CRFCA in fiscal 2027 and 2028:

• Migration, $30 million to $40 million a year. The study-stage work and the 23-month bridge extension the NRO gave HawkEye in December 2025 roll into CRFCA at roughly today's rates. This is the floor case, and it already justifies the vehicle.

• Expansion, $55 million to $70 million a year. End-user demand pushes the first option year higher once fiscal 2027 money arrives. On the NRO's own tasking curve, this is the natural outcome.

• Platform, $100 million a year or more. CRFCA becomes the intelligence community's standing RF layer, alongside the electro-optical commercial imagery contracts. A supplemental bill or a classified plus-up gets it there.

Each step up the ladder is worth several dollars a share, and none of it is in the current price. Since the award, the NRO has named no second RF production vendor, no addendum and no new study tranche. HawkEye's base year is protected.

IV. Allied navies are buying through the US Navy

After the NRO, HawkEye's largest visible US government relationship is the Navy's NIWC Pacific in San Diego, which buys HawkEye RF data for a maritime domain awareness program across the Indo-Pacific. We pulled every task order.

The current vehicle is a sole-source IDIQ, N66001-25-D-0065, with a $98.9 million ceiling and an ordering period to 2030. $23.9 million has been ordered so far, which leaves about $75 million of room to grow inside a contract that is already awarded.

The more interesting finding is who pays. The federal procurement record flags foreign funding, and across both Navy vehicles 97.5% of HawkEye's $53.4 million of lifetime obligations is Foreign Military Sales money: Indonesia, Malaysia, the Philippines, Thailand and Vietnam on the earlier contract, State Department counter-narcotics funding, and now India.

India is the big one. The Defense Security Cooperation Agency notified Congress in April 2025 of a $131 million Foreign Military Sales case for Indian maritime domain awareness, with HawkEye named as principal contractor. HawkEye's task order under it carries a $71.1 million potential value through 2031.

For investors, this is diversification that does not show up in the headline numbers. A meaningful part of HawkEye's "US" revenue is allied navies voting with their own defense budgets, routed through the most trusted buyer in the world. It is demand that keeps flowing whatever happens to US appropriations.

V. Japan: the dedicated-capacity deal

HawkEye's anchor customer outside the United States is Japan. The evidence points to one channel. HawkEye's related-party revenue matched Japan's share of revenue exactly in 2024 and 2025, 26% and 16%. Sumitomo Corporation has been a HawkEye investor and "strategic partner" since 2018, and the former CEO of Sumisho Aero-Systems now runs HawkEye Japan. Japan's Defense Intelligence Headquarters pays Sumisho Aero-Systems ¥3.77 billion, about $25 million, for fiscal 2026 "radio wave situation acquisition services."

In December 2025, HawkEye signed a $100 million dedicated-capacity contract with an unnamed strategic partner. It pays $5 million a quarter from the second calendar quarter of 2027 for five years, and that schedule lines up exactly with Japan's fiscal years 2027 to 2031. Our work points to Japan on eight independent lines of evidence, including a Japan-based Senior Solutions Architect role posted in August and the orbital geometry of HawkEye's newest cluster (Section VI).

Dedicated capacity is a different product from data resale. The customer reserves satellite time over its own areas of interest. It is stickier, it is non-cancellable, and it is the model that built the sovereign businesses at Planet and ICEYE.

The open question is whether the dedicated deal sits on top of the existing resale business or replaces it. Japan's August budget request showed zero on the resale line for fiscal 2027. That has happened before: the fiscal 2023 request was also zero, and the line was funded at ¥1.92 billion. The decisive document is the Cabinet budget draft in late December. If the resale line comes back, or a new multi-year commitment appears, Japan becomes worth roughly $40 million a year to HawkEye from 2027, nearly double today, and our model adds $1.50 to $2.50 a share.

VI. The fleet is about to get bigger

HawkEye's product is only as good as its clusters, so we pulled orbital data for every HawkEye object from CelesTrak and read the company's FCC filings.

Ten three-satellite clusters, Clusters 5 to 14, hold formation today and do the work. Cluster 7 has demonstrably reboosted, which shows the newer satellites can maintain their orbits for years. The older clusters are retiring on schedule: Cluster 3 will re-enter around the turn of the year, and Cluster 2 has lost formation. In our model, losing Cluster 3 costs less than one formation pass a day over any area we tested, because the clusters above it already cover the same ground.

What matters is what is being added. Two new clusters launch this month: Cluster 15 on SpaceX's Bandwagon-5 rideshare, targeting October 19, and Cluster 16 on a Rocket Lab Electron from New Zealand, no earlier than October 21. HawkEye's in-house Block 3 "Kestrel" satellite, which management says costs far less per cluster to build, follows in early 2027. The FCC license already authorises 20 clusters and up to 174 satellites through 2033, so there is room to keep building.

We propagated the current fleet plus the new clusters for seven days over 14 areas of interest, counting a pass only when all three satellites in a cluster are above 20 degrees elevation at once. A mid-inclination Cluster 15 cuts the median gap between formation passes over Japan from 28 to 21 minutes and over Korea from 25 to 19. With Cluster 16 and Kestrel added, the Strait of Hormuz drops from 37.5 minutes to 21.5, and the Taiwan Strait from 37 to 23.5.

Faster revisit is what sovereign customers pay for. It is also why we think the dedicated-capacity customer is Japan: a mid-inclination orbit spends most of its time near 40 to 45 degrees latitude, which is Tokyo and Seoul.

VII. The moat

HawkEye is not alone in commercial RF anymore, and it does not need to be. Here is where the field stands:

• Spire holds a $14 million Air Force Research Laboratory contract to fly a four-satellite RF formation through 2028. It is a study and demonstration award, not a production contract.

• ICEYE and Sierra Nevada sit on NRO study tranches worth about $0.3 million to $1.2 million each.

• Unseenlabs of France is building for the French sovereign program, CASSIOPÉE.

• BAE Systems' Azalea cluster has been in orbit since November 2025 with UK customers.

• China's Diting-01 launched three satellites in September and cannot sell to the US or its allies.

Against that, HawkEye has ten working three-satellite formations, the only production RF contract at the NRO, a $131 million Indian program with its name on it, years of tasking history with US combatant commands, and ITAR rules that make it very hard for a foreign buyer to own. Our estimate is that HawkEye still captures roughly 70% to 85% of visible commercial space-RF revenue. Papers from China's National University of Defense Technology try to reproduce HawkEye's three-satellite geolocation method. Adversaries study the leader.

VIII. A balance sheet that buys time

HawkEye has about $496 million of net cash and no debt after repaying its term loans from IPO proceeds. That is about $4.54 a share, roughly 35% of the market value.

It burns cash, as every space company at this stage does. First-half operating cash flow was a positive $8.4 million, helped by a collection from the Japanese partner and customer prepayments. Stripping those out and subtracting satellite spending, underlying free cash flow was about minus $33 million for the half. At that rate the cash covers the build-out several times over, with no need to come back to the market. That matters: in our study of space IPOs that fell 50% below offer, the ones that recovered were the ones that never had to raise equity at the lows.

IX. Who is selling, and who has to buy

The IPO lock-up ended in September, and venture holders gain demand registration rights around November 2. That has been described as a cliff. The evidence says otherwise.

• No venture selling. Six weeks after the lock-up, Form 4 filings show no sales by Insight Partners, NightDragon or Razor's Edge.

• Holders are sitting on gains. We estimate Insight's blended cost at around $7.50 a share. Seraphim Space Investment Trust marked its stake at 199% of cost in June. Holders with gains and no pressure tend to wait for better prices.

• The mechanics are slow. A demand registration needs a majority of registrable holders to ask, and the company can defer it twice by 60 days.

• Insiders are holding. The Chief Information Officer exercised 13,729 options on October 14 and sold none. The CEO's sales have been small and pre-scheduled under a 10b5-1 plan.

Then there is the buyer the market has not priced. HawkEye's index float was tiny at IPO. Under FTSE Russell's rules, the post-lock-up float should qualify it for the Russell 2000 at the December reconstitution: rank day October 30, preliminary lists November 13, effective after the close on December 11. We estimate passive index buying of roughly 5 to 7 million shares, about $60 million to $85 million, with about 80% probability. That is five or more days of average volume arriving on a known date.

X. The price ladder

Here is how we think about value, step by step from the floor. Every rung is either a market reference or our own model, labelled as such. The model values HawkEye 12 months out as fiscal 2027 revenue times an EV-to-revenue multiple, plus net cash adjusted for a year of cash flow, on a treasury-method diluted share count.

• $4.54: net cash per share. The hard floor.

• $8.90 to $10.40: takeout of the business as it stands. Defense signals and electronic warfare deals have run at 2 to 3 times revenue.

• $12.72: peer median. BlackSky, Planet, Spire and Redwire trade at a median of about 3.29 times 2027 revenue. HawkEye at $12.97 trades right on it, with faster growth than most of the group.

• $14.53: re-rating on the base case. Fiscal 2027 revenue of $225 million with CRFCA as migration and Japan flat, at 5.0 times revenue, the multiple the Street pays for defense data names growing 40%-plus.

• $15.50 to $19: strategic takeout. Lockheed Martin is already an investor and integration partner. CACI, L3Harris, Anduril, BAE and Leonardo DRS are logical buyers. ITAR and classified access keep foreign bidders out.

• $16.82: bull case. CRFCA expansion, Japan additive and the European defense ministry contract of up to $75 million converting. Fiscal 2027 revenue of $270 million at 5.0 times.

• $19.45: CRFCA as a platform. CRFCA at $100 million a year or more lifts fiscal 2027 revenue to about $300 million, at 5.5 times.

• $21.87: platform plus Japan additive. About $320 million of fiscal 2027 revenue at 6.0 times.

• $22 and $23: Piper Sandler and BofA targets. Piper initiated at Overweight on October 15.

• $26: the IPO price.

• About $30.70: the average of the ten analyst targets we track.

• $35.73: the 52-week high.

• $41: the highest Street target.

For balance, we also run a bear case of $7.85, which assumes a guide-down in November, Japan's dedicated deal replacing the resale line rather than adding to it, CRFCA flat and further slips. Weighting all three of our core cases, our blended model sits at about $10.85. We show it because our job is to show the whole distribution. What makes the setup interesting is how many dated catalysts can move it inside the next ten weeks: two launches, an index inclusion, a funded CRFCA order and a Japanese budget line. The market is pricing the stock as if none of them land.

XI. What could go wrong

A bull case is only worth reading if it names its risks.

• Legacy US revenue. It slipped 13% in the first half. If CRFCA does not show funded volume by early 2027, the bears are right about the US business.

• Japan replacement. If the December budget draft confirms the dedicated deal replaces the resale line, Japan is roughly flat from 2027 rather than doubling.

• Backlog quality. Of $292 million of backlog, $105 million is contractually committed, and $100 million of that is the Japan-shaped dedicated contract starting in 2027. The rest can be cancelled or is unordered ceiling, which is normal for a government contractor but leaves the second-half guide dependent on new orders.

• A second CRFCA vendor. The NRO says it will "onboard new providers as their technologies mature." Spire is the obvious candidate.

• Execution. HawkEye's older roadmaps slipped often. We score its 2015 to 2022 promises at 13 met or early against 34 late or missed, and we haircut every roadmap date by six to twelve months.

• ISA. HawkEye is holding back $17.5 million of the ISA purchase price against a Justice Department civil investigative demand over ISA's cybersecurity compliance. We found no unsealed complaint.

XII. The catalyst calendar

The next ten weeks hold more dated, binary catalysts than most companies see in a year.

November 12, Q3 earnings. Consensus revenue is $56.5 million (eight analysts, $55.0 million to $57.9 million). The October 15 announcement of $36 million in new international contracts lands mostly in backlog, which sets up the fourth quarter. Here are the ten things we will be listening for and what resolves each one bullish.

Six or more bull resolutions on that list and we expect the stock to start climbing the ladder in Section X.

HawkEye is a real company with a real technical lead, a balance sheet that carries it through the build-out, a customer base that is visibly using more of its product every year, and a production contract its competitors do not have. The market is pricing it as a failed IPO. The filings describe a company that has just been handed the vehicle to scale. The gap between those two pictures is what the next ten weeks will close.

In Part 2 we will go inside the satellites and the unit economics: what a Block 2 cluster costs, what Kestrel saves, how much revenue each working cluster earns, and the revenue level at which this business turns sustainably profitable.

Research, not investment advice. Shawarma Capital has no position in $HAWK.

References

• HawkEye 360 Q2 2026 Form 10-Q: https://www.sec.gov/Archives/edgar/data/1750704/000175070426000022/hawk-20260630.htm

• HawkEye 360 SEC filings index: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001750704

• House Armed Services Committee hearing print, National Security Space Programs, May 14, 2025 (NRO questions for the record): https://www.congress.gov/119/chrg/CHRG-119hhrg61761/CHRG-119hhrg61761.pdf

• Japan Ministry of Defense budget requests, FY2027 / FY2026 / FY2024 / FY2023: https://www.mod.go.jp/j/budget/gaisan/r9/gaisanyoukyu.pdf , https://www.mod.go.jp/j/budget/gaisan/r8/gaisanyoukyu.pdf , https://www.mod.go.jp/j/budget/gaisan/r6/gaisanyoukyu.pdf , https://www.mod.go.jp/j/budget/gaisan/r5/gaisanyoukyu.pdf

• USAspending award records, Navy IDIQ N66001-25-D-0065 and task orders: https://www.usaspending.gov/award/CONT_IDV_N6600125D0065_9700

• DSCA Transmittal 25-22, India Indo-Pacific Maritime Domain Awareness: https://www.dsca.mil/Press-Media/Major-Arms-Sales/Article-Display/Article/4170460/india-indo-pacific-maritime-domain-awareness

• SBIR award 209979, Spire FA9453-24-C-X046: https://www.sbir.gov/awards/209979

• FCC grant DA 24-262 (HawkEye 360 constellation): https://docs.fcc.gov/public/attachments/DA-24-262A1.pdf

• CelesTrak HawkEye 360 orbital elements: https://celestrak.org/NORAD/elements/gp.php?NAME=HAWK&FORMAT=tle

• Intelligence Community News, NRO awards CRFCA contract to HawkEye 360: https://intelligencecommunitynews.com/nro-awards-crfca-contract-to-hawkeye-360/

• Umbra, NRO radar contract: https://umbra.space/blog/umbra-nro-radar-contract/

• Breaking Defense, NRO extends contracts for four providers (December 2025): https://breakingdefense.com/2025/12/nro-extends-contracts-for-4-providers-readies-new-acquisition-approach/

• HawkEye 360, $36 million in new international contracts (October 15, 2026): https://www.prnewswire.com/news-releases/hawkeye-360-secures-36-million-in-new-international-contracts-driven-by-increasing-demand-for-signals-intelligence-302909323.html

• HawkEye 360, Indian Ocean Region maritime domain awareness contract (July 1, 2026): https://www.prnewswire.com/news-releases/hawkeye-360-awarded-multi-year-contract-to-enhance-indian-ocean-region-maritime-domain-awareness-302832888.html

• Piper Sandler initiation (October 15, 2026): https://www.investing.com/news/analyst-ratings/piper-sandler-initiates-hawkeye-360-stock-with-overweight-rating-93CH-4950874

• Insight Partners Schedule 13G (August 14, 2026): https://www.sec.gov/Archives/edgar/data/1305473/000089914026000889/primary_doc.xml

THIS IS NOT FINANCIAL ADVICe

Read more

Disclosure: Shawarma Capital has no position in $PL, long or short. This is research, not a recommendation.

Planet Labs reported $116.1 million of revenue for the quarter ended July 31, 2026. It was the largest quarter the company has ever had, 58% above the same quarter a year earlier, and it beat the top of its own guidance. The stock still fell, and it has kept falling: from the June 4 print to October 2, $PL is down about 59%, against 44% for BlackSky and 27% for the UFO space ETF over the same window.

The obvious explanation is that the whole space trade cooled after the SpaceX listing in June. That is part of it. The other part is sitting in Planet's own 10-Q, and it is less obvious. The record quarter was assembled from contracts signed in earlier years, and the order book that is supposed to feed the year after next got thinner while it happened.

This piece walks through that arithmetic, then follows the money to the customers behind it, then counts what Planet actually has flying. It ends with the strongest bull case we can build, a dated list of tripwires, and our valuation model, clearly labelled as a model.

I. What the record quarter borrowed

Planet does not report bookings. It does report backlog, a non-GAAP measure that adds cancelable contract value to remaining performance obligations, and that is enough to back the number out.

Backlog was $906.1 million at April 30, 2026. It was $814.9 million at July 31, 2026. In between, Planet recognised $116.1 million of revenue. So net new contract value added in the quarter was about $25 million, a backlog book-to-bill of roughly 0.2×.

That is our calculation, not a company figure, and it is a net number: it includes any cancellations, scope changes and currency moves. Strip out roughly $8 million to $10 million of FX and it is closer to $33 million to $35 million, about 0.3×. Second quarters are seasonally soft for Planet, and the comparable figure a year earlier, excluding the big sovereign deals, was lower still. On a trailing-twelve-month basis with the three disclosed sovereign mega-deals removed, book-to-bill has slipped from about 1.05× a year ago to somewhere between 0.7× and 0.95× now, depending on the undisclosed size of the Swedish contract (about 0.9× if it is near $110 million).

The backlog decline itself is not a secret. Zacks wrote on September 23 that backlog "fell 10% sequentially", and at least two other outlets made the same point in early September. What has not been mapped is where in time the decline landed.

Planet's 10-Qs say what share of backlog it expects to recognise within 12 and 24 months. At April 30, 2026, those figures were approximately 40% and 69%. At July 31, 2026, they were approximately 50% and 70%. Multiply through and the slice due in the next 12 months rose to about $407 million. The slice due in months 13 to 24 fell from about $263 million to about $163 million, a drop of roughly 38% in one quarter. Because the company's percentages are rounded, the honest range is a fall of 33% to 43%. The direction does not depend on the rounding.

The fair reading needs to be stated before the bear reading. Using the rounded percentages, roughly $60 million to $80 million more moved into the near-term window than a straight-line roll would predict. Some of that is almost certainly lumpy satellite work, where a handover scheduled for mid-2027 crosses the 12-month line, and some may be accelerated milestones. The filings do not say. They also do not rule out reductions in later-year scope. What they do show is that the year-two book was not refilled. At the October 2025 Investor Day, management held up $177 million of committed year-two backlog as a proof point. The equivalent figure at July 31, 2026 is about $163 million.

One more piece of the quarter deserves a footnote. About $14.1 million of Q2 revenue was point-in-time hardware: the handover of a Pelican satellite to the Swedish Armed Forces. The company had guided that handover into Q3. The CFO said on the September call it was "shifting revenue between the 2 quarters without changing our full year outlook." Excluding it, Q2 revenue was roughly $102.9 million to $103.9 million against a guide of $102 million to $107 million. By our count it is the first quarter in six without an underlying beat.

The FY28 math

That year-two slice is where FY28 revenue (the year to January 2028) has to come from. As of October 2, 2026, sell-side consensus for FY28 is $570.6 million from 12 analysts, with a range of $539 million to $605 million. That implies about 31% growth on the midpoint of Planet's FY27 guide.

Our backlog bridge, which assumes Planet adds new business at last year's pace, lands at roughly $465 million to $565 million, with a central case near $500 million to $530 million. A separate Monte Carlo that prices every renewal, option and pipeline item we could identify (200,000 draws) puts FY28 at a median of $497 million, with an 80% interval of $452 million to $554 million. In that model, consensus sits near the 95th percentile.

The reason is recognition speed. Planet's sovereign deals do not turn into revenue quickly. In the first full year after signing, the JSAT contract contributed about 7.6% of its $230 million value. The German-funded €240 million contract contributed roughly 9% to 18%. Sweden contributed 27% to 40%, but only because an already-built satellite was handed over. At a blended 17%, closing the gap to consensus needs roughly $560 million of new satellite-services contract value signed by January 31, 2027. In the last twelve months Planet signed about $140 million of that kind of work (Sweden plus Germany's civil BKG framework).

Management's answer is the pipeline. On the September call the CEO described "over $4 billion of identified opportunities for satellite services, over 25% of which is qualified as near-term pipeline," and the CFO said near term means "quarters, not years." That is the bull case in one sentence, and we return to it in section V.

II. Who actually pays: Ukraine, through Berlin

Planet's FY26 10-K contains a sentence we could not find quoted in any English-language market coverage: "EMEA includes revenue from Ukraine of $35.9 million for the fiscal year ended January 31, 2026." That is 11.7% of FY26 revenue of $307.7 million. Ukraine and Japan ($38.0 million) are the only two countries other than the US named above the 10% line.

Planet's own description of its largest European deal never mentions Ukraine. The July 1, 2025 release announced "a multi-year €240 million agreement, funded by the German government." GAU (deaidua.org) was the outlet that reported on June 9, 2026 that Ukraine is the end user of that contract. The primary document behind that report is a written answer from the German government to the Bundestag, Drucksache 21/4186, question 46, answered on February 19, 2026.

The answer is short and specific. No tender took place. There is no contractual relationship between Planet and the Federal Republic of Germany. The €240 million contract is between Ukraine and Planet, and Germany financed it from its Ukraine military-aid budget line (Einzelplan 60, Kapitel 6002, Titel 687 06) as an ex-gratia payment under §53 of the Federal Budget Code. The year-by-year breakdown the questioner asked for was not given.

This matters for three reasons.

Concentration. Defense and intelligence was 69.8% of Q2 FY27 revenue, and most of it is not the US government. Our estimate, from filed geography and concentration data, is that D&I was roughly 72% foreign in Q2. Three unnamed customers accounted for 17%, 11% and 10% of Q2 revenue. Planet does not name them, and the filings do not prove Ukraine is a single customer, so we do not call it one.

Backlog share. We estimate the Ukraine contract is roughly a fifth to 30% of Planet's $815 million backlog (about 21% to 31%), and the three sovereign satellite-services deals together (Ukraine via Germany, JSAT in Japan, and Sweden) roughly 60% to 75%. Planet does not disclose backlog by customer. These are estimates with wide error bars.

Tail risk. The budget line that pays for the contract is large and, for now, stable: €11.55 billion in 2026 and €11.61 billion in the 2027 draft. But the payment is discretionary by definition, and a ceasefire would change Kyiv's priorities and Berlin's. Our base case puts Ukraine-attributed revenue at roughly $41 million to $56 million over the next twelve months. A scenario in which a ceasefire cuts that by a quarter removes more revenue than Planet's entire FY27 adjusted EBITDA guide of $3 million to $10 million. We do not think that is the most likely path. It is a path the current valuation does not visibly price.

Planet's Berlin subsidiary, by its own accounts

Planet Labs Germany GmbH files audited accounts, and the most recent set is attached to its entry in the Bundestag lobby register. For the year to January 31, 2025, before the €240 million deal was announced, it booked €46.6 million of external revenue, all from Council-of-Europe states. It paid €38.2 million of licence fees to the US parent, about 82% of that external revenue. EBITDA was €3.0 million, a 4.8% margin on total revenue of €61.9 million. It carried €38.0 million of deferred revenue.

On June 19, 2026, commercial-register filings for the GmbH replaced the group's President and CFO as managing director and changed the company's registered purpose from geographic information services to include the manufacture and operation of satellites and ground stations. Three months later Planet opened a Berlin factory it says can scale to "up to 60 satellites per year."

Germany's own sovereign-constellation award to Planet, as opposed to the Ukraine contract it finances, is a five-year framework for the federal mapping agency BKG with a ceiling of €25.2 million net of VAT (€30 million gross), signed on August 4, 2026 after an open procedure with two bidders.

The JSAT relationship has its own cash fingerprint. SKY Perfect JSAT's annual securities report for the year to March 2026 shows ¥16.2 billion (about $104 million) already paid on its ¥37 billion Pelican programme, most of which is Planet's $230 million contract, before any JSAT satellite has launched. First launches are expected in 2027. That is a large part of why Planet's cash flow looks better than its income statement.

Planet reported positive FY26 free cash flow of $52.9 million. Deferred revenue, mostly customer prepayments, rose $151.1 million in the same year. Without that build, free cash flow would have been roughly negative $98 million, or about negative $68 million if the part of those billings not yet collected is netted out. On the same basis, free cash flow was negative in 13 of the last 14 quarters through July 2026. Current deferred revenue stood at $281.2 million at July 31, 2026, and the 10-Q now spells out that milestone invoices "generally require customer acceptance prior to payment," including "acceptance of commissioned in orbit satellites."

III. Europe is buying radar

The bull case leans on European defence budgets, and they are real. Germany has announced €35 billion for space defence through 2030. The question is which sensor the money buys.

We built a ledger of European government Earth-observation contracts with publicly disclosed values signed between September 2024 and September 2026. It totals about €3.5 billion. Radar (SAR) took about 88% of it. ICEYE, counting its Rheinmetall joint venture's €1.7 billion SPOCK-1 contract at gross value, accounted for about 61%. Airbus, building Spain's PAZ-2, took about 26%. Planet took about 10%, or about 4% if the Ukraine-signed, German-funded €240 million contract is excluded.

Treat that as our ledger of disclosed values, not a market share. It mixes satellite builds with data and capacity contracts, it uses an estimate for Planet's Swedish deal, and it leaves out several ICEYE awards (Netherlands, Sweden, Ukraine) whose values were never published, which means ICEYE is understated. Counting SPOCK-1 at ICEYE's 40% equity share instead of gross value, ICEYE falls to about 45% and Planet rises to about 15%.

The scale of the competitor is now public. ICEYE's Finnish statutory group accounts show 2024 revenue of €103.0 million and an operating loss of €49.2 million. The company says unaudited 2025 revenue more than doubled to over €250 million, with EBITDA above €100 million. It raised money in June 2026 at a valuation of about €10 billion. Planet's 10-K and 10-Qs do not name ICEYE, Capella or Umbra as competitors.

The US picture rhymes. When the NRO handed out commercial radar awards in August 2026, they went to Capella, ICEYE US and Umbra. BlackSky's 10-Q shows its US federal revenue down 16% year on year in the June quarter while its international government revenue rose 173%. And Planet's largest customer in FY23 to FY25, at 19% to 21% of revenue (roughly $36 million to $46 million a year), was almost certainly the NRO under its Electro-Optical Commercial Layer contract. Planet does not name it. In December 2025 the NRO renewed only its PlanetScope baseline, for $13.2 million through June 2026, a narrower scope than the 2024 renewal. We estimate the NRO book at roughly $20 million to $30 million annualised now. Planet has not said whether the contract continued after June 2026; "EOCL" and "NRO" appear zero times in the Q2 10-Q and press release.

The NASA line needs a calmer reading than it has been getting. NASA has not yet placed its next PlanetScope order under its commercial data programme. The current $13.5 million order, itself 32% below the prior year's, runs to November 24, 2026. That alone is not alarming. In 2024 NASA signed Planet's renewal on the day it started, NASA's own June 18, 2026 release lists Planet among continuing contract holders, and none of last year's tasking vendors has been renewed yet. The risk is appropriations under a continuing resolution that ends December 11, not lost share.

IV. What is actually in orbit

Investor pieces published this year describe Planet's PlanetScope fleet as "150+ active" or "250 operational" satellites. Planet's own October 1 release says it has built and delivered "718 satellites" on orbit, which is a cumulative figure. The public catalogue tells a narrower story.

CelesTrak's satellite catalogue, pulled on October 2, 2026, shows 110 Dove-family satellites still in orbit, down from a peak of 357 in February 2022, a fall of 69%. That peak count includes every Dove generation, among them older Dove-Classic satellites that were no longer working. Of today's 110, CelesTrak flags 92 as operational or partly operational. Counting only the current SuperDove generation, satellites in orbit fell from 185 in January 2024 to 109 today, a drop of 41%.

Two caveats cut in Planet's favour. Planet launched 18 more SuperDoves on October 1, 2026. They are not yet catalogued and most were still on a carrier spacecraft as of October 2; if all of them commission, the working count goes to about 110. And Planet's own "operational" label means the satellite can be manoeuvred, not that it is imaging; the company says the status is "not meant to provide any indication with respect to the tasking capabilities."

Two caveats cut the other way. By our arithmetic, a gapless once-a-day pass over the equator needs about 80 SuperDoves, or about 95 with modest overlap. The 92 that CelesTrak flags as working cover about 1.15 times the equator per day. That is a geometric floor, not proof of coverage gaps, but it says Planet has been running close to the minimum, and the October 1 launch is needed to keep margin. Second, the satellites are not lasting as long. SuperDove cohorts launched in 2019 to 2021 stayed in orbit a median of about 3.6 years. Those launched in 2022 and 2023 lasted about 2.6 to 2.7 years, with the solar maximum a large part of the cause.

The annual reports moved with the fleet. Planet's first 10-K as a public company said it had "approximately 200 satellites" in orbit. The 10-Ks for FY23 to FY25 said "hundreds of satellites" in orbit. The FY26 10-K dropped "in orbit" and now says Planet has "built, launched, and operated hundreds." CelesTrak flags 117 Planet-built satellites of all types as operational or partly operational today.

Pelican: the promise and the spec sheet

In April 2022 Planet said Pelican, its high-resolution tasking satellite, would "begin launching early next year" and consist of "up to 32" satellites imaging at "up to 30 cm" and "up to 12 times per day." As of October 2, 2026, Planet says ten Pelican-family satellites are on orbit. Two are technology demonstrators. One, which Sweden calls EDDA-1 and Planet's own documents identify as Pelican-7, was handed to the Swedish Armed Forces. That leaves seven Planet-owned production satellites.

Planet's May 2026 Pelican product specification gives the current satellites a native panchromatic resolution of 77 cm at a 440 km reference altitude, 54 cm after processing, resampled to 50 cm in delivered products. The fleet flies at 443 km to 489 km, which by simple scaling puts native resolution nearer 0.77 to 0.86 metres. Planet's 10-K and 10-Q filings have never claimed better than "up to 50 centimeters after processing." The 30 cm figure appears only in press releases, including two earnings releases furnished to the SEC in June 2022 and June 2026, and there only as a design goal.

This would be a footnote if the next generation were not carrying so much contract value. Roughly $510 million of signed sovereign work, the $230 million JSAT contract and the €240 million German-funded contract, is tied to Gen-2 Pelicans, which Planet says are "designed to provide up to 30 cm class imagery." That capability has not yet been shown in orbit. Pelican-12, the second Gen-2 satellite, is manifested on SpaceX's Bandwagon-5 rideshare, no earlier than late October.

Behind Pelican sits Owl, Planet's planned successor to the SuperDove, described in releases as "near-daily, 1 meter class" imagery. Owl is never named in Planet's 10-K or 10-Qs. It appears in two press releases furnished to the SEC. The periodic reports refer only to "medium resolution satellites" under construction and warn that next-generation satellites "may not be completed on time or at all." In September, Planet raised its FY27 capex guide to $100 million to $115 million, which the CFO described as "investments in Pelican and Owl."

On the March 2025 call, management called FY26 "the peak of the growth CapEx investment cycle." FY26 capex came in at $81.5 million, 26.5% of revenue. The FY27 guide is 23% to 27% of revenue, against a long-term target of 5% to 8%. The SuperDove successor is being paid for before its first demonstration has flown, and Planet's launch path for small satellites depends on SpaceX rideshares that, according to trade press reports SpaceX has not confirmed, are not taking new commercial bookings beyond late 2028.

V. The Google line items

The rally on October 2 coincided with Planet's October 1 launch of a prototype satellite for Google's Project Suncatcher, the space-based compute experiment. The accounting is worth knowing before treating it as a revenue story.

Under an October 2025 R&D agreement with Google to test "a payload integrated on prototype satellites," which we believe is Suncatcher (the word appears in no SEC filing), Google's payments reduce Planet's R&D expense rather than count as revenue. Through July 31, 2026, Planet had recognised $5.2 million of that offset against $9.5 million of cash received.

Google is also Planet's cloud landlord. Planet's discounted Google Cloud terms run to January 31, 2028. Absent a new deal, the contract defaults to Google's then-current list prices; $48.0 million of minimum commitments remained at July 31, 2026. A December 2024 amendment was never filed as an exhibit, so we cannot rule out that it changed the term. Our estimate of the list-price step, if nothing is renegotiated, is $6 million to $18 million a year of extra cost, set against a FY27 adjusted EBITDA guide of $3 million to $10 million.

Google was once a customer too. Its content-licence revenue to Planet, mostly the run-off of a 2017 prepayment, fell from $11.9 million in FY23 to $0.3 million in FY25 after the licence ended in November 2024. Google holds about 35.25 million Planet shares, roughly 10.6%, and has launched imagery-embedding products of its own.

VI. The bull case, built as strongly as we can

A balanced piece owes the other side its best argument, so here it is.

The sovereign model works, and it is fast. Planet signed three satellite-services contracts in twelve months (January 2025 with JSAT, July 2025 with Ukraine via Germany, January 2026 with Sweden), worth over half a billion dollars together. Sweden signed on January 12, 2026, its first satellite launched on May 3 and was handed over in the July quarter. The CEO says Planet delivered first satellites "within 2 and 4 months of the contract award." No listed peer has shown that speed. Two of the three prior deals were announced in late January, which is the next seasonal window.

The pipeline and the hiring point the same way. Management cites $4 billion of identified opportunities, about $1 billion of it near term. On September 29, 2026, two new ground-system stacks appeared in Planet's certificate logs under its on-premises "Direct Access Service" domain; one is a full ground node whose customer, if there is one, is not public. Planet had 121 open roles on October 2, the most since November 2022, including NATO pre-sales roles in the Netherlands and Belgium, Korea and Stockholm deployment roles, two UAE roles and two "Constellation Services" marketing roles.

The balance sheet can carry the wait. Cash and short-term investments were $865.4 million at July 31, 2026, about $405 million net of the $460 million 2030 convertible notes. Interest income ran at $6.4 million in Q2, roughly $25.6 million a year, more than the whole FY27 adjusted EBITDA guide. Planet sold $122.4 million of stock through its at-the-market program at about $32 a share and has not reported sales below that level.

The industrial base is coming online. Berlin opened on September 22. Isar Aerospace's Spectrum reached orbit on September 5, which removes the "unflown rocket" objection to the Berlin-built Pelican planned on Isar in early 2027. Pelican-12 is the fifth Pelican launch of 2026, against five in 2025.

Adoption is still compounding where it is measurable. Academic works mentioning PlanetScope rose from 212 in 2019 to 1,689 in 2025, and the ratio to Sentinel-2 works roughly doubled to about 10%, flat in 2026 rather than falling.

Our probabilities on the bull case: a new satellite-services contract of $100 million or more announced by January 31, 2027, about 40%; by January 31, 2028, about 70%. On the bull lane's own assumptions, the chance FY28 revenue reaches consensus is about 30%. Our Monte Carlo, which prices each line separately, puts it at 5.7%, and plausibly 4% to 10% once correlated US budget outcomes are allowed for. The gap between those two numbers is the trade. The bull case is credible and funded. It is also an event bet on converting a self-reported near-term pipeline that has not produced a nine-figure contract in the 8.7 months since January 12.

VII. What to watch, with dates

Each item below is a dated event that moves our numbers. We will update the model against each one.

• Through October 30: the convertible-note trigger window. Immaterial: the notes settle principal in cash and a capped call covers dilution up to $18.04.

• No earlier than late October: Pelican-12 (Gen-2) on Bandwagon-5. The test is the first-light resolution claim and whether it says native or resampled.

• About November 5: BlackSky Q3. Its US federal revenue line is the only public quarterly read on NRO and NGA commercial imagery spend.

• Early to mid November: JSAT's Japan Ministry of Defense optical-data follow-on (last year's was announced November 4). Absence by late November would be a real signal.

• November 12 and November 27: the Bundestag budget committee's final session and the final vote on Germany's 2027 budget. Check the Ukraine-aid title, Einzelplan 60, Kapitel 6002, Titel 687 06.

• About November 16 to 18: Planet announces its Q3 date.

• November 24: NASA's current PlanetScope orders ($13.5 million plus $0.9 million) end. Watch whether peers' renewals appear before Planet's.

• October to December: NATO's SINBAD Step 2 analytic-alert award, where Planet is the incumbent from Step 1.

• About December 8 to 10: Q3 FY27. The numbers that matter: backlog (at or above $815 million means book-to-bill of 1× or better), Q3 revenue at or above about $107.6 million (which keeps the $430 million low end easy), the point-in-time revenue line, satellite inventory, APJ revenue as a read on the JSAT build, and current deferred revenue.

• December 11: the US continuing resolution expires.

• Late January 2027: the historical window for satellite-services announcements.

• February 1, 2027: the evergreen equity-plan increase, about 22 million shares at the full 6%.

• No earlier than February 2027: Transporter-19, the likely ride for further Swedish Pelicans.

• About March 18 to 26, 2027: Q4 FY27 and the first FY28 guide. A first guide in the low $500 millions would read as a large miss against $571 million.

• About May 24, 2027: the NRO EOCL five-year base period ends.

What would change our mind: a nine-figure satellite-services contract signed before January 31, 2027 that recognises revenue quickly (an in-orbit handover structure rather than a multi-year build), a Q3 book-to-bill at or above 1×, or Gen-2 imagery that is natively near 30 cm.

VIII. Valuation, as model output

Everything in this section is the output of our model, not a forecast or a price target.

At about $17.6 on October 2, 2026, $PL trades at 13.8 times enterprise value to the FY27 revenue guide, 10.5 times FY28 consensus and 7.4 times backlog. BlackSky, growing at a similar rate and already guiding to positive adjusted EBITDA of $12 million to $24 million, trades at 6.3 times forward sales and 2.3 times its RPO. Fitting a line through the listed Earth-observation peers (BlackSky, Spire, Satellogic, MDA), Planet carries about a 2.1× premium to where its growth would place it. The one comparable that supports the current price is ICEYE's private round, at roughly 40 times trailing sales for a radar company with an EBITDA margin of about 40%.

Our scenario model (200,000 draws, 12% cost of equity) gives these 12-month values:

• Stress (7%): no new mega-deal plus a Ukraine, budget or US-restriction shock. About $6.28.

• Bear (25%): no new mega-deal and one of those shocks. About $6.66.

• Base (46%): FY28 revenue near $508 million, multiple compresses as growth slows. About $11.13.

• Bull (22%): a new sovereign deal lands, Gen-2 works and Ukraine continues. About $16.24.

The probability-weighted 12-month value is $10.84, or $9.68 in today's money, against a price near $17.6. The model gives about an 8% chance that present value exceeds the current price. A sum-of-the-parts cross-check lands at $5.49 to $12.02 a share.

Two conditional results matter more than the headline. If a new sovereign deal of $150 million or more lands by January 31, 2027, the mean present value is about $13.10. If not, it is about $7.41. And if FY28 lands exactly at consensus with FY29 growth of 25% and today's premium held, the model gives $16.83, still below the current price. In other words, the stock is priced for consensus plus a sustained premium, and consensus itself needs bookings Planet has not yet signed.

Share count is the quiet input. Class A and Class B shares outstanding rose 18% in the twelve months to August 27, 2026, from 307.9 million to 363.85 million, mostly from warrant exercises, an earnout tranche and the at-the-market program. $1.378 billion of at-the-market capacity remains.

Bottom line

Planet is a real business with a real sovereign product, a strong balance sheet and a management team that has shown it can ship satellites to a government customer in months. None of that is in dispute. What the July quarter shows is that the record revenue was delivered out of a year-two order book that was not refilled, that roughly 70% of revenue now comes from defence and intelligence buyers who are mostly foreign governments, that one of the largest is Ukraine with Germany paying, and that the fleet behind the data franchise is smaller and shorter-lived than most investor coverage assumes. The Q3 print in December will either show the bookings that consensus needs or confirm the gap. We have no position, and we would want to see that print before having one.

References

• Planet Labs PBC Form 10-Q, quarter ended July 31, 2026: https://www.sec.gov/Archives/edgar/data/1836833/000119312526382016/pl-20260731.htm

• Planet Labs PBC Form 10-Q, quarter ended April 30, 2026: https://www.sec.gov/Archives/edgar/data/1836833/000119312526258304/pl-20260430.htm

• Planet Labs PBC Form 10-K, fiscal year ended January 31, 2026: https://www.sec.gov/Archives/edgar/data/1836833/000119312526119957/pl-20260131.htm

• Q2 FY27 earnings release, Exhibit 99.1 (September 3, 2026): https://www.sec.gov/Archives/edgar/data/1836833/000119312526381874/pl-ex99_1.htm

• Q1 FY27 earnings release, Exhibit 99.1 (June 4, 2026): https://www.sec.gov/Archives/edgar/data/1836833/000119312526257401/pl-ex99_1.htm

• Q1 FY23 earnings release, Exhibit 99.1 (June 14, 2022): https://www.sec.gov/Archives/edgar/data/1836833/000183683322000034/ex991.htm

• Germany €240M agreement release, Exhibit 99.1 (July 1, 2025): https://www.sec.gov/Archives/edgar/data/1836833/000183683325000075/exhibit991projectchronospr.htm

• Swedish Armed Forces agreement release, Exhibit 99.1 (January 12, 2026): https://www.sec.gov/Archives/edgar/data/1836833/000119312526009922/pl-ex99_1.htm

• Q3 FY26 earnings release, Exhibit 99.1 (December 10, 2025): https://www.sec.gov/Archives/edgar/data/1836833/000119312525314160/pl-ex99_1.htm

• Google Cloud agreement exhibits 10.18 and 10.20: https://www.sec.gov/Archives/edgar/data/1836833/000119312521297137/

• Deutscher Bundestag, Drucksache 21/4186, question 46 (answered February 19, 2026): https://dserver.bundestag.de/btd/21/041/2104186.pdf

• GAU, "Germany funds €240 million satellite deal for Ukraine" (June 9, 2026): https://deaidua.org/news/2026/06/09/germany-funds-240-million-satellite-deal-ukraine/

• Planet Labs Germany GmbH annual accounts, year to January 31, 2025 (Bundestag lobby register R007542): https://www.lobbyregister.bundestag.de/media/82/71/850484/Planet-Labs-Germany-GmbH_JAP_31-01-25_Offenlegung-1.pdf

• Commercial register mirror, Planet Labs Germany GmbH, HRB 175210 B: https://www.northdata.com/Planet+Labs+Germany+GmbH,+Berlin/Amtsgericht+Charlottenburg+(Berlin)+HRB+175210+B

• TED notice 576400-2026, BKG sovereign satellite constellation award: https://ted.europa.eu/en/notice/576400-2026/xml

• Bundestag 2027 budget timetable: https://www.bundestag.de/dokumente/textarchiv/2026/kw34-haushalt-2027-ablauf-1187766

• SKY Perfect JSAT annual securities report, fiscal year to March 31, 2026 (EDINET S100YBNS): http://www.kabupro.jp/edp/20260616/S100YBNS.pdf

• Swedish Government release, January 12, 2026: https://www.regeringen.se/pressmeddelanden/2026/01/over-53-miljarder-kronor-till-bred-dronar--och-rymdformaga/

• Swedish Armed Forces release, January 12, 2026: https://www.mynewsdesk.com/se/forsvarsmakten/pressreleases/foersvarsmakten-satsar-paa-rymdfoermaagor-3424270

• Rheinmetall and ICEYE SPOCK contract release (December 18, 2025): https://www.rheinmetall.com/en/media/news-watch/news/2025/12/2025-12-18-rheinmetall-and-iceye-win-billion-euro-contract-for-space-reconnaissance

• ICEYE group financial statements via PRH (Taloustutka): https://www.taloustutka.fi/konserni/2639822-1

• ICEYE 2025 financials release: https://www.iceye.com/newsroom/press-releases/iceye-announces-2025-financials

• ICEYE valuation, Via Satellite (June 9, 2026): https://www.satellitetoday.com/finance/2026/06/09/iceyes-valuation-soars-to-10-billion-euro-with-new-raise/

• NRO commercial SAR awards, Via Satellite (August 5, 2026): https://www.satellitetoday.com/government-military/2026/08/05/capella-iceye-us-and-umbra-receive-nro-commercial-sar-contracts/

• BlackSky Form 10-Q, quarter ended June 30, 2026: https://www.sec.gov/Archives/edgar/data/1753539/000175353926000124/bksy-20260630.htm

• BlackSky Q2 2026 earnings release: https://www.sec.gov/Archives/edgar/data/1753539/000175353926000118/exhibit991-blackskyq22026e.htm

• NASA contract release C26-007 (June 18, 2026), mirror: https://www.globalsecurity.org/space/library/news/2026/space-260618-nasa01.htm

• USAspending.gov, NASA orders 80HQTR25FA066 and 80HQTR25FA012: https://www.usaspending.gov/

• CelesTrak satellite catalogue (pulled October 2, 2026): https://celestrak.org/pub/satcat.csv

• Planet operational status file: https://ephemerides.planet-labs.com/

• Planet, "Next Generation Satellite: Pelican" (April 2022): https://www.planet.com/pulse/next-generation-satellite-pelican/

• Planet Pelican Imagery Product Specifications (May 2026): https://planet.widen.net/s/kbt9m6p6sx/planet-userdocumentation-pelicanproductspec

• Planet release, Pelican-12 arrives at launch site (September 25, 2026): https://www.stocktitan.net/news/PL/planet-s-pelican-12-satellite-arrives-at-launch-vquiqv1hjkdj.html

• Planet release, Berlin manufacturing facility (September 22, 2026): https://www.stocktitan.net/news/PL/planet-opens-state-of-the-art-satellite-manufacturing-in-yygbdgvmac83.html

• Planet release, Suncatcher, Tanager-2 and 18 SuperDoves launched (October 1, 2026): https://stockhouse.com/news/press-releases/2026/10/01/planet-launches-suncatcher-tanager-2-and-18-superdove-satellites

• Isar Aerospace second flight release: https://isaraerospace.com/press/history-for-european-spaceflight-isar-aerospace-reaches-orbit-and-deploys-payloads-on-second-flight

• Satnews on SpaceX rideshare bookings (July 20, 2026): https://satnews.com/2026/07/20/spacex-limits-future-commercial-rideshare-bookings-past-2028-amid-severe-manifest-saturation/

• Zacks via Yahoo Finance on Planet backlog (September 23, 2026): https://finance.yahoo.com/markets/stocks/articles/planet-labs-turn-backlog-visibility-142700249.html

• Yahoo Finance analyst estimates for PL (captured October 2, 2026): https://finance.yahoo.com/quote/PL/analysis/

• Q2 FY27 earnings call transcript (September 2026): https://www.fool.com/earnings/call-transcripts/2026/09/09/planet-labs-pl-q2-2027-earnings-call-transcript/

• Greenhouse job board API, Planet: https://boards-api.greenhouse.io/v1/boards/planetlabs/jobs

• Certificate transparency log search: https://crt.sh/?q=%25.das.planet.com

• OpenAlex works search, PlanetScope: https://api.openalex.org/works?search=%22PlanetScope%22&group_by=publication_year

Disclaimer. This publication is for informational and educational purposes only and is not investment advice, a recommendation, or an offer to buy or sell any security. Shawarma Capital holds no position in Planet Labs PBC ($PL) as of October 2, 2026. Figures described as our calculations, estimates or model outputs are derived from public filings and data and may be wrong; company filings are the authoritative source. Do your own research.

SHMD closed the Friday before that piece at $6.00. It closed yesterday, the twenty third of September, at $3.68.

So the line aged well. But it was a line about a category, and a category is not a company. We had not read a single SCHMID filing when we wrote it. This piece is what happens when you do.

We did not stop at the filings, because this is a company where the filings are the least reliable instrument available. SCHMID reported €66.9 million of revenue for 2025 after guiding to "up to €110 million". It projected €130 million for 2024 when it listed and delivered €60.8 million. It cut its 2026 margin guidance thirty nine days after repeating it in a registration statement. When a management team's numbers move that much, you stop listening to the numbers and go and weigh the machines.

That is literally what we did. A SCHMID wet-process line is a physical object. It weighs somewhere between three and sixty tons. It is built in one of two plants, one in the Black Forest and one in Guangdong, crated, declared to customs under a tariff code the company itself types onto the shipping documents, loaded onto a ship, and delivered to a factory that usually has to file for a permit to install it. Every one of those steps leaves a record that management does not write.

So we pulled 42 United States bills of lading naming Gebr. SCHMID as shipper. We pulled 44,089 rows of monthly European export statistics and matched individual shipments to individual machines by weight and month. We ran a census of every land auction in the Zhongshan district where SCHMID says it is building a new campus. We read the Korean regulatory filings of its joint venture partner, the Taiwanese stock exchange filings of thirty two of its potential customers, the Chinese state tender results, the Air Force Research Laboratory's delivery address, a Michigan economic development grant, 243 patent families, two earnings call transcripts, every SEC comment letter, a New York federal docket and the property records behind its own headquarters.

What we found splits cleanly into two findings that do not fit together, and the rest of this piece is an attempt to hold both of them at once.

Two findings that do not fit together

The first finding is that the business is finally working. SCHMID booked €96.6 million of equipment orders in the first eight months of 2026, more than in all of 2025. Its backlog of €95.0 million on the twenty first of August is the highest figure it has disclosed since 2022. First-half revenue of €46.0 million was up 172 percent. The Chinese plant is, in the CFO's words, "pretty much booked out". The orders are coming from the part of the electronics supply chain that AI servers and optical transceivers are stretching hardest: high-layer HDI boards, mSAP lines, flip-chip substrates and panel-level packaging. Customs records, export statistics and customer filings independently confirm that machines are moving.

The second finding is that the equity is still a financing. The share count rose from 37,974,862 at the listing to 61,637,954 on the twenty second of September, a 62 percent increase, almost none of it for cash at a premium. The company's convertible noteholder converts at a discount to recent trading prices, with a floor of $1.93, and has done so eleven times, usually on or right after an up day. A Chinese state-linked fund that received 12.5 million shares at $2.15 to settle a debt sold 2.8 million of them between February and May at an average of about $6.49. The chief executive sold 1.9 million shares in the four trading days after the company's biggest press release of the year. Total equity was minus €132.2 million at the end of 2025. It has published twenty two press releases since the listing, several of them market-moving, that it did not furnish to the SEC when it issued them.

Both findings are true. The mistake most people make with a stock like this is to pick one and ignore the other. Bulls see the backlog and assume the equity will follow. Bears see the dilution and assume the backlog is fiction. We think the backlog is not fiction and the equity will not simply follow, and the gap between those two statements is where the trade lives.

Read more

Nothing about the business changed in those forty eight hours. What changed was the venue, the underwriter, and the story. That gap between the venue and the business is the entire subject of this piece, and once you understand how it opened, everything else about this company becomes legible.

On the thirtieth of April, Bleecker Street Research published a short report calling SharonAI a neocloud built on phantom contracts and questionable financing. The stock was around $63 at publication. It is $50 now. The short is working, and most people have concluded the report was simply right.

We spent three research rounds and sixteen agents testing that conclusion against primary sources. We pulled the 555 page Indian prospectus of the customer at the centre of the report. We pulled the bankruptcy docket the report cites. We pulled the Form D that names the underwriter's fee. We read the audited accounts of the CEO's private vehicle in Australia. We enumerated the company's certificate transparency logs, its autonomous system number, its PeeringDB facility registrations and its status page incident history.

What we found is that the report was right about the paperwork and wrong about the price. Its two load bearing claims have not survived. Several of its sharpest lines are materially imprecise. And the single most serious conflict at this company is one the report did not identify at all.

The thesis in two sentences

SharonAI is a real, very small, asset light GPU cloud with roughly two to four megawatts actually connected, wrapped in promotional numbers its own filings do not support, and the market has already discounted the promotional numbers to approximately zero. At $50 the enterprise value is $1.18 billion against $1.86 billion of cash, which means cash is 94 percent of the market capitalisation, only $29.60 a share is at risk on the operating business, and two independent valuation back solves both land on the capacity figure in the audited filings rather than the one in the press releases.

That is not a defence of the company's disclosure. It is an observation that you are being asked to pay for the conservative case and handed the rest as optionality.

1. What this company actually is

Strip away the language and SharonAI is a tenant.

It does not own data centres. It sold the one it had. The flagship Texas campus, a 250 megawatt joint venture in the Permian Basin called Texas Critical Data Centers, was a fifty fifty partnership with a small helium and natural gas company. SharonAI sold its entire fifty percent stake in January 2026, one month before the Nasdaq pricing, for $70 million. It was paid early and in full, taking $74.8 million, and booked a $65.9 million gain. That gain, plus $5.0 million on the shares it received, is the only thing that made money in the first half of 2026.

This matters for a reason that is not obvious. The company carries $26.3 million of property and equipment against $2.32 billion of total assets. That looks alarming until you know the data centre was sold and the GPUs are not owned either. The fleet sits in two leased halls belonging to NEXTDC, the Australian colocation operator, in West Footscray Melbourne and Artarmon Sydney. A thousand GPU Blackwell cluster is consumed under a Lenovo pay as you go agreement, meaning it never touches the balance sheet at all. A further $302.65 million sits in prepayments to named suppliers, principally $256.19 million to WWT Australia, for hardware that has not been delivered.

So the accounting is not hiding anything. The accounting is telling you precisely what this is: a company that has paid cash to vendors for equipment it has not received, to install in buildings it does not own, for customers whose contracts begin in 2027.

The cloud itself is real, and we had to correct ourselves on this point. An early pass concluded the company was a brochure site with no product. That was wrong, because it probed the wrong domain. The actual platform lives on a fourth domain registered in August 2025, with working single sign on, self serve registration, 352 pages of technical documentation covering managed Kubernetes, GPU virtual machines, serverless inference and per second metering, and production hostnames resolving inside the company's own announced address space. The company holds a genuine autonomous system number registered to its founders' pre existing Australian vehicle, announcing 1,536 addresses across six blocks.

It is a real cloud. It is also a small one. The company's own status page, live since October 2025, lists two regions, both Australian, and records three incidents in eleven months. Self reported peering traffic is twenty to fifty gigabits per second. A production cloud carrying meaningful load generates considerably more of all three.

2. The short report, adjudicated

We took twenty six discrete claims from the Bleecker report and tested each against primary documents. Thirteen confirmed. Six overtaken by events. Five overstated. Two false. Two unresolved.

The pattern is consistent and worth stating plainly, because it is to the author's credit. Where Bleecker quoted a document, it quoted correctly. The failures are inferential rather than documentary. This was not a sloppy report. It was a report written in April about a situation that moved in May, June, July, August and September.

What died

The centrepiece was ESDS Software Solution, an Indian company that signed a $1.25 billion five year contract with SharonAI on the thirty first of March. The report's argument was arithmetic and looked devastating: ESDS earned $39.9 million and held $69.5 million of assets, yet had committed to pay roughly $250 million a year and post $140 million in letters of credit, a figure exceeding its entire balance sheet.

We pulled ESDS's own Red Herring Prospectus, filed with the Indian securities regulator on the twenty fourth of August. It is 555 pages, audited, and legally binding on its signatories. It did not exist when Bleecker wrote.

It confirms the contract term for term. Entered the thirty first of March 2026. Five years plus a two year option. Total contract value of $1.25 billion, stated in rupees as ₹118,312.50 million. Approximately 8,208 NVIDIA B300 GPUs, a more precise figure than SharonAI's own "8K". An Australian data centre. Monthly service fees.

The phantom contract claim is false. The contract exists, and the customer's own auditors have examined the disclosure.

More importantly, the prospectus reveals the funding mechanism, which nobody outside the two companies could previously see. ESDS's subsidiary has already received an advance of ₹11,766.35 million, approximately $124.3 million, from an enterprise customer incorporated outside India, for this project. That advance is what tripled ESDS's balance sheet, taking cash from ₹607 million to ₹12,534 million.

This is a prefunded back to back resale. ESDS is not a small company promising to pay a quarter of a billion dollars a year out of its own earnings. It is an intermediary that has already collected a large advance from the ultimate end user. And the number ties to something unexplained on SharonAI's own balance sheet: a customer deposits line of $143,879,911, carrying no note, no policy and no counterparty disclosure. The two sides of the trade reconcile.

The second load bearing claim was financing. The report attacked a $500 million facility from USD.AI, a decentralised finance protocol, noting correctly that it had roughly $284 million of lending capacity against $1.2 billion of approvals.

Every word of that arithmetic was right, and it remains right. We pulled USD.AI's live proof of reserves data. Its entire loan book across twenty seven deals sits in Sweden, Canada, France and the United States. Australia is zero. New Zealand is zero. Its first accountant attestation, published on the first of September, states the book is deployed across facilities in the United States, Canada and Sweden.

SharonAI never drew a dollar. Every filing for eight months has carried the identical sentence, that USD.AI had approved a facility subject to execution of definitive documentation. It was never upgraded to "we entered into" or "we drew". USD.AI appears zero times in the second quarter report.

The report named a lender that was never used. And in the interim the company raised approximately $2.15 billion from institutions that are emphatically not decentralised finance protocols: Oaktree, which holds $375 million across both convertible tranches, plus Millennium, Soros, Point72, Blackstone, Diameter, Sona and Lone Pine. Cash went from $71 million to $1,861,347,822. Annual interest is $54.6 million against 34 times coverage, and nothing matures before May 2031.

Three further claims decayed. The assertion that ESDS cannot legally serve Indian customers from Australia rests on Indian data localisation rules that cover payment system data only, and the end customer here is offshore, so the rule never bites. The Gazprombank exposure was confirmed to the decimal for fiscal 2025 at 20.15 percent, and has since fallen to 2.80 percent. The supervoting structure described as sixty five percent of votes on one percent of economics was a mismatched pairing then, and today stands at 38.2 percent of votes against 45.7 percent of economics, which is not a voting majority at all.

What survived, and what got stronger

The governance critique stands, and in one respect we strengthened it against the company.

The DSS acquisition, in which the founders sold a private Australian entity into the public shell, drew a comment from the Securities and Exchange Commission itself during the merger review, questioning whether it was a common control transaction that should carry no goodwill at all. Bleecker did not have that. It is corroboration from the regulator.

The NVIDIA misstatement is confirmed and remains live. The annual report filed on the thirty first of March, signed by the chief executive, stated that NVIDIA is a strategic shareholder in SharonAI. An 8-K filed on the thirteenth of April stated NVIDIA holds no equity securities. The annual report has never been amended, so the false sentence is still in the filing today, and the correction was filed under the "other events" item rather than the non reliance item. In fairness, the sentence sits in a single cell of a partner table whose adjacent row makes the identical claim about a different partner and is true, which reads as a copy into the wrong row rather than an invention. It is sloppy. It is not fabricated. It is also uncorrected six months later.

And the letters of credit do not exist. ESDS's certified indebtedness table, examined by an independent accountant as at the thirtieth of June, three months after signing, records letters of credit sanctioned nil and outstanding nil. Its entire non fund based limit is approximately $5.5 million against a required $140 million. The customer's own legally binding disclosure document does not record the security that SharonAI's lenders are being asked to underwrite.

What was overstated

Three lines in the report do not survive contact with the underlying record, and anyone repeating them should stop.

The report says the chief executive's prior public company is suing him for routing $11.5 million through a shipping company he controlled. There is no such case. A full sweep of all thirty Mawson Infrastructure dockets, plus searches on his name and his vehicles, returns zero federal actions naming James Manning as a party. Across 120 docket entries in the live adversary proceeding, his name and his company's name appear zero times. The allegation lives in Mawson's answer to an involuntary bankruptcy petition, a defensive pleading offering motive evidence against the creditors who filed it, and it has never appeared in any Mawson annual or quarterly report. The word identifying that shipping company returns zero hits in both.

The payments that did flow to that company from SharonAI were $167,638 in 2024 and $92,722 in 2025, and it appears nowhere in the 2026 related party note.

The report describes a $25 million self dealing transaction in DSS. The audited accounts of the CEO's own vehicle show it sold DSS for A$1,018,000 of stock, booking a A$900,074 gain on a carrying value of approximately A$118,000, which ties precisely to the $677,000 SharonAI disclosed. The $25 million was the whole company valuation across mostly unrelated holders. The insider take was about one million Australian dollars.

And the contractor arrangement the report attacked, under which the chief executive, chief financial officer and a co founder invoiced through their own companies, was terminated in April 2026 with no notice period and no payment.

On the regulatory record the negative is clean and deserves to be stated as plainly as the allegations. No action by the Securities and Exchange Commission. No banning or disqualification by the Australian regulator. No Nasdaq proceeding. No industry record. No criminal charge. No restatement, no non reliance filing, no auditor resignation over any of it. After three years of allegations, no regulator anywhere has acted.

These are serious, specific and entirely untested allegations. The fair weight is a standing governance discount, because the related party density is undisputed disclosure rather than allegation, plus a guidance credibility discount. It is not a fraud thesis.

3. What nobody has reported

Here is the part of this piece that is ours, and it is the most serious conflict at this company.

The underwriter that took SharonAI public bought its stock at $6.00 first.

The sequence is documented in two filings. A prospectus supplement filed on the thirteenth of February shows Lucid Capital Markets as one of three investors in $2,250,000 of December 2025 bridge notes that converted at $6.00 a share. Six days later Lucid acted as representative of the underwriters on the public offering priced at $30.00. On the seventeenth of April it initiated research coverage with a buy rating and a $50 price target. Nine days after that, a Form D filed on the fourth of May names Lucid Capital Markets as the sole recipient of sales compensation on the $350 million convertible offering, paid $14,875,000.

Bought at six. Sold to the public at thirty. Published a fifty target. Collected fourteen and a half million.

Lucid is placement agent on every financing this company has done, roughly $2.075 billion in total. And its Head of Capital Markets was the sole officer and director of the SharonAI registrant before the merger, and is himself registered to resell 465,343 shares. That same individual joined the board of the shell in 2023.

Which brings us to the shell. SharonAI did not arrive on Nasdaq through an initial public offering in any meaningful sense. The vehicle began as TKB Critical Technologies 1, which raised money in October 2021 to buy a connected car data company. That deal collapsed. In June 2023 the sponsor and directors sold 4,312,500 shares and 8,062,500 private warrants to affiliates of Roth Capital Partners and Craig-Hallum for one dollar. The vehicle was renamed, then failed two separate Nasdaq listing rules, one for having fewer than four hundred holders and one for not holding an annual meeting. It voluntarily delisted in April 2024, liquidated its trust, and saw ninety percent of its public shares redeemed.

That is the entity SharonAI merged into in December 2025, landing on the OTC Pink market. Two months later it uplisted to Nasdaq at $30.00, two days after trading at $1.90.

We also found a selling shareholder that bought at $6.00 whose trust belongs to the chairman of both the CEO's private investment vehicle and the shipping company named in the Mawson pleading. None of that relationship appears in the selling shareholder footnotes.

Two more items nobody has published.

The company bought land in Australia and did not tell its shareholders. On the ninth of September a regional newspaper reported that Visy sold its former paper mill near Albury, New South Wales, to SharonAI, with a company spokesperson confirming it on the record, corroborated by a second outlet and by a statement from the local member of parliament. This is the first evidence of any owned real property. It appears in no filing and in no press release. There is no development application lodged. It is bare dirt, disclosed to a regional paper and an MP but not to the market.

And the headline backlog includes the supplier. The second quarter release presents $8.8 billion of total contract value, and books $4.9 billion of that under a heading called customer momentum, attributed to NVIDIA. But the financial statements say the company committed to procure hardware for approximately $4 billion, and the chief executive describes the NVIDIA pricing as guaranteed as a minimum only, a floor rather than a ceiling, focused on the resale of that capacity. That is a purchase obligation and a price backstop presented as customer demand. Signed customer contract value is $3.893 billion, not $8.8 billion.

4. The capacity question, which is the real one

The company says 212 megawatts. Its own quarterly filing says 132 megawatts secured and 116 contracted. The number 212 appears nowhere in the filing's body.

We traced its origin. It first appears on the sixth of August, in an exhibit to the second quarter results. And contrary to our own earlier working note, it is not a contradiction of the release two days before. On the fourth of August the company said total capacity remains at 132 megawatts with 120 contracted. On the sixth it added a previously unannounced additional 80 megawatts, bringing the total to 212.

So a sixty one percent increase in claimed capacity was disclosed as a bullet in an earnings release and a remark on a call. No standalone filing. No facility named. No host named. No contract. Approximately 80 of the 212 megawatts has no disclosed home anywhere in the public record.

Against that, what is actually connected today is two to four megawatts, in two NEXTDC halls, verifiable through the company's own PeeringDB registration, which lists exactly two facilities and zero in the United States. The fleet was 432 GPUs as of September 2025, including sixteen H200s the company disclosed were acquired primarily for customer latency testing and proof of concepts.

The honest way to describe the gap is not fiction. It is tense. This is a 2027 business being described in the present tense. The largest tranche, 72 megawatts at NEXTDC worth roughly $623 million, does not deploy until April to August 2027. Three of the four headline contracts begin in 2027. The quarterly filing concedes the point in its own language, recording no liability because services have not commenced.

There is real corroboration that the contracts are binding. NEXTDC's own exchange filing on the twenty first of July reports contracted utilisation rising 73 megawatts in the same month SharonAI signed a 72 megawatt order. That is a landlord confirming the lease from the other side.

5. The tape, and why it fell

We began this work with a hypothesis that the decline from $97 to $50 was mechanical, driven by convertible arbitrage funds shorting the stock to hedge their bonds. We were wrong, and the data killed it cleanly.

Peak short interest was 2,744,284 shares, 7.8 percent of the class, against a theoretical fully hedged book of about 7.5 million. Short interest fell throughout the entire decline, from 2.74 million in July to 2.36 million at the end of August. Days to cover ran between one and one point four. Borrow is easy and cheap, with 800,000 shares available at 2.45 percent. The stock has never appeared on the threshold list. The options book is 3.3 to one long calls, which is retail buying upside, not desks hedging bonds.

The actual explanation is supply, and it is dated to the day.

Class A shares went from 11,832,164 at the end of December to 35,667,164 by August, roughly tripling in six months. On the fourteenth of August, 13,094,719 shares priced at $68.73 became freely saleable. On the twenty first, a further 8,056,699 priced at $12.53 joined them. That is 21,151,418 shares, 59.3 percent of the entire class, unlocked in eight days, on registration statements the Securities and Exchange Commission expressly declined to review. The stock fell 23.2 percent in the four sessions between the two effective dates, and the high of $76.90 came three sessions after the first.

Seventy six percent of that freed equity is unhedged.

The largest holder is worth understanding precisely, because it is simultaneously the biggest risk and a meaningful signal. Situational Awareness, the fund run by the former OpenAI researcher Leopold Aschenbrenner, holds 8,070,950 shares and prefunded warrants, 21.1 percent on the basis its own filing uses. Its cost is $523,882,863, or $64.91 a share, leaving it roughly $105 million underwater. Its fund shrank from $45 billion to $10 billion. It converted from a passive to an active filing on the twenty eighth of August, and the nineteen point nine nine percent ownership blocker ceased to apply on the twenty seventh.

It has sold nothing. Zero shares. Through a forty six percent drawdown.

Neither have insiders, in either direction. Five ownership forms since listing, one of them a zero dollar grant, no sales, no plans, no buying.

And the retail story is a non story. The company did run a paid investor relations campaign, disclosed properly under the relevant rule, costing $200,000 in total, $50,000 in monthly fees and $150,000 for a four week television campaign that aired in March. It lapsed in the second quarter. The promotional window and the price window do not overlap. A census of 1,356 messages on the main retail forum returns no coordinated phrasing, and the twelve and a half percent decline on the fourteenth of September drew four messages. Retail has left. It did not make this price, and it is not defending it.

6. What you are actually paying for

This is where the argument turns, and it turns on arithmetic rather than opinion.

At $50.17 the basic share count is 35,803,505 and the treasury method count is 39,430,551. Enterprise value is $1,180.9 million, and across six different definitions the range is $0.83 billion to $1.39 billion, narrow enough that the choice does not matter.

Cash is 94.1 percent of the market capitalisation. Only $29.60 a share is exposed to the operating business at all.

Now run the two back solves.

At the peer median of roughly $9.0 million of enterprise value per megawatt, $1,180.9 million implies 131 megawatts. The quarterly filing says 132.

At the peer median of 3.48 times forward revenue, it implies $339 million of 2027 revenue. That is 44 percent of the signed book's run rate and 22 percent of the four analyst consensus.

Two independent methods, approached from different directions, both land on the number in the audited filing rather than the number in the press release.

The 212 megawatts is not in the price. The $8.8 billion is not in the price. The 68,000 GPUs are not in the price. Every promotional figure this piece has spent six sections dismantling has already been discounted to approximately zero by the market. The promotional record bear case is paid for.

Meanwhile the company trades at 0.77 times forward revenue against a peer median of 3.48, the cheapest in its sector by a factor of 3.2, while simultaneously trading at $328 million per live megawatt against a peer band of $45 to $64 million. Both are true at once, and the reconciliation is the cash.

The unit economics are also better than the narrative implies. The contracts price at $18.6 to $19.0 million per megawatt per year, which is below one listed peer's disclosed figure and inside another's stated band. The pricing is market. The delivery is the anomaly. Gross margin is driven far more by depreciation life than by utilisation: on a four year life the range across seventy to one hundred percent utilisation is negative 21.5 to positive 12.9 percent, while on a six year life it is positive 13.2 to positive 37.1 percent.

And the funding shortfall, correctly framed, is a growth gap rather than a solvency gap. Total uses through 2027 are $2,141.6 million against $1,861.3 million of cash. That is short $280.3 million with no customer prepayments, short $111.3 million at ten percent, and a $57.7 million surplus at twenty percent. We know prepayment is real, because ESDS has already advanced $124.3 million. Interest coverage is thirty four times and nothing is due before May 2031.

7. The dated tests, and what they are saying right now

This thesis is unusually falsifiable, which is the best thing about it. Four tests carry dates.

The first was the sixteenth of September, the contractual delivery date for the ESDS cluster. As of writing there is no filing, no announcement, and no evidence anywhere that the $140 million security has been posted. More pointedly, on the tenth of September the customer's own chairman told Indian television that the Sydney facility would go live in the next forty to fifty days, by the end of October. Six days before the contractual date, the customer was publicly guiding six weeks past it. He did not name SharonAI, and the connection runs through the only Australian arrangement ESDS discloses, so treat it as strong inference rather than proof.

The second is the twenty sixth of September, when a 15 megawatt Sydney facility was said to be going live. The landlord's last public statement was in May. Its newsroom has not posted since June and has never named SharonAI. The company's PeeringDB registration still shows two facilities, last updated on the twenty ninth of July. No third building has appeared.

The third is secured financing. On the twenty first of August the convertible holders consented to strip the covenants limiting debt and liens, which is unambiguous preparation for asset backed borrowing, and which Oaktree as a required holder with a veto necessarily approved. Twenty five days later nothing has been filed.

The fourth is the cleanest and cannot be spun. When the next quarterly report is filed, either it recognises an operating lease liability for the new capacity or it does not. Accounting standards require recognition when a lease commences. If those megawatts are live, the liability appears. If it does not appear, they are not live.

Two of four have failed so far. One is pending with the customer guiding past it. That is the honest scoreboard, and it is why this is a 2027 story.

8. What would make us wrong

If the September and October deliveries land and the next quarterly report shows a lease liability stepping up, the live footprint roughly quadruples and the 2027 book starts converting to revenue. That is the bull path and it is measurable.

Against it, here is how the long thesis breaks.

The ESDS contract is renegotiated or shrinks, which is the single most likely adverse outcome given no letter of credit exists, the customer is a reseller with a twenty percent gross margin, and its own analyst models an earnings margin collapsing from 50.8 to 18.1 percent.

The secured financing does not arrive, or arrives at punitive terms, leaving the 2027 tranche unfunded. Watch for any filing that follows the covenant strip.

Situational Awareness sells. Eight million shares, twenty one percent of the class, a hundred and five million dollars underwater, in a fund that lost most of its assets, with the blocker now removed. Nothing obliges it to be patient.

The May convertible strikes at $48.24, which is at the money today. A recovery arms that dilution rather than relieving it.

The auditor position deteriorates. Three firms in twenty months, no internal control audit ever performed, and a material weakness in complex financial instruments that remains unremediated.

Or delivery simply keeps slipping, and 2027 becomes 2028, at which point the cash cushion that makes this defensible starts being consumed by interest and overhead rather than by building.

The position

The short report was right about the paperwork. The capacity number is promotional, the backlog counts the supplier, the console has been removed from public DNS, the land purchase went to a regional newspaper instead of the market, and the underwriter bought at six and sold you thirty.

It was wrong about the price. The contract is real and prefunded. The financing it attacked was never drawn, and two billion dollars arrived from Oaktree and Millennium and Soros instead. The lawsuit it described does not name the man it describes. The twenty five million was one million. And the market has already marked this company at the number in its audited filing rather than the number in its press release.

At ninety four percent cash backing, you are paying roughly thirty dollars a share for a business the market is valuing at the 132 megawatts on file, with the 2027 book as free optionality and four dated tests in the next ninety days to tell you whether it converts.

That is not a story stock at this price. It is a financing and delivery question with a large cash cushion behind it, and the honest answer is that we will know a great deal more by the end of October than we know today.

Disclosure and disclaimer. This is our own research and opinion. It is not investment advice and not a solicitation to buy or sell any security. Figures are drawn from company filings, exchange filings, foreign regulatory prospectuses, court records and public infrastructure registries, and we make no warranty as to their accuracy or completeness. Bleecker Street Research has disclosed a short position in SHAZ. Do your own work.

A subscriber asked the right question in the chat this week. What are the nearest catalysts, and when do we expect them. Another said he felt overdue for some bigger news. Both are fair, and both deserve a straight answer instead of a pep talk.

So this …

Read more

Read more

Read more

One note before we start, because it matters: Merlin hosted this visit, and everyone I met was careful, professional, and disciplined about what a public company can and cannot say. Nobody gave me projections, timelines, financial information, or anything you cannot find in a filing. What I brought home was not information. It was calibration: the chance to check the public record against the physical reality. Every program fact, number, and quote in this piece comes from public filings, press releases, and the recorded management call previously published to subscribers. The rest is what I saw with my own eyes and heard through my own headset.

There is a specific kind of due diligence you cannot do from a terminal. I have spent five months on this company from the outside: the filings, the FAA registry, the lobbying disclosures, the flight trackers, the federal budget documents, thirteen parts of it published here. I have tracked their airplanes across two hemispheres and read their certification milestones the way other people read box scores. What I had never done, until this week, was sit inside the product while it worked.

So when the invitation came to come to Rhode Island and fly, I took it.

Quonset

To get to Merlin's flight test operation you drive onto the old Quonset Point naval air station in North Kingstown, Rhode Island, and the geography does half the storytelling before you ever reach the hangar.

This is the place that gave the Quonset hut its name. The Navy built the base in 1941, the Seabees were born here, and for a generation this stretch of Narragansett Bay was one of the busiest military airfields on the East Coast. The first thing you pass today is General Dynamics Electric Boat, where they assemble hull sections for nuclear submarines: cranes, security fencing, barge docks, the whole industrial choreography of the most complicated machines America builds. Quonset is not a tech campus with a climbing wall. It is a place where the government has been betting on hard, unforgiving engineering for eighty years, and the neighborhood tells you what category of thing is being attempted inside.

The second thing you notice, if you know what to look for, is that the Rhode Island Air National Guard's 143d Airlift Wing flies C-130Js off this same field. The airframe at the center of Merlin's defense business takes off and lands, routinely, on the other side of the fence from Merlin's hangar. Nobody staged that for my benefit. It is simply where the company chose to do its work: submarines on one side, Hercules on the other, and in between, a hangar where a software company is teaching airplanes to fly themselves.

Merlin moved flight test here from Mojave in late 2024 to be near the Boston headquarters, and the hangar houses the Cessna Grand Caravans that anchor the FAA supplemental type certificate program, the civil half of this story. I parked in the hangar lot, walked in past the caution tape and the toolboxes and the aircraft in work, and the first impression was the important one: this looks exactly like what the filings say it is. A working flight test unit in the middle of a campaign. Not a showroom. Nobody had tidied the lab for company.

The room

The company put real people in front of me, not a deck. The VP of Investor Relations. Flight test staff. Another investor, a PM at a hedge fund, was visiting the same day, which tells you something about who is doing primary work on this name right now. On the management call last week, the CEO mentioned that Merlin had been extending these invitations for a while and only recently had investors started taking them up. Having now gone, I understand why the ones who do come back different.

We sat down for a pre-flight meeting and talked for a long while about the intricacies of aviation and autonomy, and I want to describe the texture of that conversation, because texture is what you fly to Rhode Island for. Nobody pitched me. The conversation kept drifting, the way conversations among competent people do, into the genuinely hard parts of the problem: what it means to put a machine in command of an aircraft in the national airspace system, how you bound a decision-maker you did not hand-write, what safety actually means when you engineer for it rather than argue about it. Everyone I met was conspicuously qualified, and everyone talked about the same thing from a different angle. The angle was never the stock. It was always the airplane.

The frame they operate under is the one the CEO has laid out publicly, and the consistency between the executive suite and the hangar floor is itself a data point, because message discipline that deep is rarer than you would think. Aviation has had assisted autonomy for decades: autopilots, autothrottles, autoland. Each of those is a tool a pilot wields. What Merlin is building is the jump from assistance to a full stack pilot, a system that owns the mission from takeoff through touchdown, perceives, decides, communicates, and acts, with the human moving up one level, to supervision and authorization. Matt George has publicly described Merlin Pilot as a student pilot: you teach it the airplane, then you expand what it is trusted to do, gradually, gate by gate. Not a feature. A crew member.

The man in the left seat

The pilot assigned to our demo flight is a former military C-130 pilot who now flies as a test pilot for Merlin. Sit with that for a second, because the resume alone carries information. The program Merlin is trying to win is crewing down the C-130J for Special Operations Command, from two pilots to one. The person in the left seat of my demo has flown that mission the old way, with two humans up front, and now spends his working life supervising the software designed to hand one of those seats to a machine.

But the resume was not the striking part. The striking part was the conviction.

Understand what a test pilot is. A test pilot is a professional skeptic. The entire job, the culture, the training, is organized around distrust: fly the machine to the edges of what it claims it can do, find where the claims break, write it up, do it again. Test pilots do not gush. They are the immune system of an aviation company. So when I sat across from one before the flight and listened to him talk about Merlin Pilot, what I was listening for was hedges, and what I heard instead was a man who has watched this system through hundreds of hours of the least charitable scrutiny available and come out the other side convinced. He talked about safety the way career aviators talk about safety, which is to say constantly, unprompted, from every direction: the discipline of the test program, the layers between an idea and an action, what the system will simply refuse to do. Not marketing safety. Procedural safety. The kind with paperwork.

His conviction reads even better against the public backdrop. The CEO has said publicly that the vast majority of aviation accidents remain the direct result of human error, and that safety is the first thing autonomy changes. A machine does not get tired on the sixth leg of the day. It does not get distracted, or rushed, or complacent, and it flies the aircraft the same way at 2 a.m. as at 2 p.m. The men and women who have spent careers watching human factors erode airmanship are precisely the people you would expect to feel strongest about that, and, at least in the sample I met, they do. When people ask me what conviction inside a company looks like, it is not the founder's conviction that impresses me. Founders are paid to believe. It is the test pilot's.

The airplane is a laboratory

Then we walked out to the ramp and inspected the aircraft, and the romance got replaced by something better, which is evidence.

Start with the airplane itself, because the choice of airframe is strategy. The Cessna Grand Caravan is arguably the most consequential utility aircraft of the last forty years. FedEx funded part of its development in the 1980s because it wanted a cheap, rugged feeder for packages; more than 3,000 of the Caravan family have been delivered since; and its PT6A turbine is close to the most trusted engine in aviation. Caravans haul cargo, skydivers, mail, and medicine on every continent. It is the world's workhorse, which is exactly why you would pick it as the first certification target: certify autonomy on the airplane that already does the unglamorous flying of the global economy, and the market for the resulting product is not hypothetical. It is FedEx feeders and island freight runs and every operation that runs these airframes hard, every day, at thin margins.

Now the part that carries more of the thesis than it first appears to: none of Merlin's aircraft are fly-by-wire. A Caravan, a King Air, a C-130J: these are cable-and-hydraulics machines designed decades before anyone imagined software in the loop. The integration process, as the CEO has described it publicly, is to retrofit servos and actuators onto the airplane, then "characterize" it, flying a set of maneuvers with the company's test pilots to teach Merlin Pilot how this specific airframe behaves, then gradually expand the envelope of what the system is allowed to do. Why that matters commercially: the world's cargo and military fleets are not fly-by-wire either. A company that can only automate new airplanes has a market measured in future deliveries. A company that can retrofit autonomy onto the existing fleet has the thousands of aircraft already out there. That is the entire Condor thesis, and it starts with servo brackets.

You see all of it in the cabin. A rack of flight test electronics with dozens of thick cables running forward, the visible circulatory system between the compute and the airplane's nerves. Test instrumentation where seats would be. And in the back, the flight test engineer's station: a second joystick, a monitor, and the tablet that runs the show.

The tablet

Our flight test engineer walked me through the backend before we ever turned the propeller, and I want to be precise about what I saw, because no press release conveys it.

The tablet is the mission. You create waypoints and place them in three dimensions, altitude included, building a route the way you would sketch one on a map that happens to be real. The aircraft flies it with complete autonomy. You can edit the plan in flight, and the aircraft accepts the new routing the moment it is authorized; there is no perceptible gap between permission and execution.

Restricted airspace renders as hard boundaries the system will not enter. Not "is trained to avoid." Will not. And the same philosophy runs all the way down the stack: the servos and actuators themselves have bounded authority, and the intelligent layer works through those bounds rather than around them. Everything the autonomy wants to do is checked against deterministic constraints before it becomes motion. If you have spent any time thinking about how a learning system gets certified for aviation, you will recognize the architecture: put the clever part inside a cage the regulator can inspect. The cage is the product as much as the cleverness is.

This is not a marketing abstraction, and here is the receipt. The two components Merlin announced it had pushed through Stage of Involvement 3 with the Civil Aviation Authority of New Zealand this month, in the joint validation program with the FAA, are exactly the two things I watched work: the Flight Control Computer and the Automated Communication System. The certification paperwork and the demo describe the same machine. That coherence, filings matching hangar, is precisely what you fly to Rhode Island to check, and it checked.

On the same screen you can watch the communication side of the brain: the ATC comms stream and the language processing pipeline, every stage of the loop from radio call to understood instruction to proposed action, laid out where an engineer can audit it. And woven through the whole walkthrough, unprompted, was redundancy: software redundancy, power redundancy, hardware redundancy. This is a company that talks like an avionics house, not like a software startup that happens to own airplanes.

The flight

We boarded, briefed, and lined up. The pilot authorized the departure. Then Merlin Pilot flew the airplane.

Not "assisted." Not "mostly." The system advanced through the takeoff, lifted off, climbed, and flew the waypoint plan. We changed the route mid-flight from the tablet and the aircraft took up the new plan instantly. It stayed inside every boundary and followed every rule, the entire time, with the two humans up front doing exactly what the architecture says they should be doing: supervising, and authorizing.

Somewhere in cruise came the moment I will be telling people about for years. Through my headset I heard a transmission go out to air traffic control, and the voice did not belong to anyone in the airplane. It was Merlin Pilot on the radio, working its own frequency, in a synthetic voice with, I have to report, genuine personality. There is something permanently mind-bending about listening to the national airspace system talk to a piece of software as a peer, and the controller on the other end not caring. This is the Automated Communication System, the same component that just cleared SOI-3: per the company's own public materials, it interprets spoken ATC instructions and converts headings, altitudes, and airspeeds into commands for the flight control system, and it reads back like any other airplane on the frequency. Reading that in a certification announcement is one thing. Hearing it key up in your headset at altitude is another.

The interface runs both directions. I was shown how a human can speak to Merlin Pilot and authorize override commands by voice, and the system will execute them if, and only if, the request fits inside its bounds. The envelope is the constitution. Everything, including the humans, operates within it.

And then, at the end of the mission, the system flew the approach and landed. No manual pilot input, from the start of the journey to the rollout. The flare was judged, the touchdown was smoother than most of the human landings I have logged as an airline passenger, and the airplane tracked the centerline like it was on a rail. Five weeks earlier, this same company flew a Caravan onto Runway 27 at EAA AirVenture Oshkosh, hands off all the way to the ground in front of the largest aviation crowd on the planet, the first autonomous landing of a conventional fixed-wing aircraft in the show's fifty-plus-year history. On our call, the CEO was asked why he would take that risk in public, and his answer was that there is a world where you do this in private, and "this is not how we operate." Having now sat behind the system while it worked a real frequency in real airspace, I believe him. The Oshkosh landing was not a stunt. It was the routine, performed in front of witnesses.

What a demo proves, and what it cannot

Here is where I put the analyst hat back on, because I am aware of the failure mode where someone gets one good demo and comes home writing a fan account.

A demo proves the system exists, works end to end in live airspace, and is mature enough that the company will put outside investors in the cabin as a matter of routine. In this sector that is genuinely rare; most autonomy stories are rendered videos and roadmap slides. But a demo is not certification, and certification is the entire civil ballgame. So let me lay the hard numbers next to the experience, because both are true at once.

Merlin's Q2 10-Q is on file as of August 14. Revenue for the quarter was $2.19 million, down 29 percent from the same quarter last year, a fact the sequential bulls skip because it rose from Q1. The revenue that exists today is defense services revenue, and certification-gated commercial revenue is still zero, exactly as the filings say it should be at this stage. Meanwhile the company is spending roughly $30 million a quarter across R&D and G&A to build the thing I flew with. What funds the gap is the balance sheet: $183.5 million in cash and short-term investments as of June 30, no debt, which at the current burn is runway into 2028 without asking anyone for money. The headline net loss figure includes a large non-cash deemed dividend from the preferred repricing; read the cash flow statement, not the EPS line.

Against that spend, the program stack as it stands, all public record:

• USSOCOM C-130J autonomy, sole prime. Critical Design Review complete June 4. Integration planning underway. The next task order is the catalyst.

• KC-135 air refueling. Teamed with GE, building on the company's prior Project Metis work. Roughly 300 tankers in the active fleet, with airframes older than the pilots flying them.

• FAA STC program, Cessna Grand Caravan. The flight test campaign at Quonset. The airplane I flew.

• NZ CAA and FAA joint validation. SOI-3 complete August 6 for the Flight Control Computer and the Automated Communication System. Commercial service target in New Zealand 2027, confirmed publicly.

• Commercial cargo, the Condor thesis. IAI and WorldStar partnerships; IAI is one of the largest passenger-to-freighter converters in the world, and the addressable fleet is measured in thousands of airframes.

• SHIELD. Classified R&D. Publicly characterized only as "interoperable collaborative autonomy."

Three things in that stack deserve emphasis after the visit.

First, the C-130J program is scoped more modestly, and therefore more achievably, than most people model. In the CEO's own words from our call: "don't forget our contract in our program, the C-130J, is to go from two pilots down to one." Not unmanned. Reduced crew. The human stays, which means the human-machine interface is the product, and the CEO has been candid publicly that this interface, the autonomy and the human pilot sitting together in one flight deck, was harder than expected and has "now turned into a deep competitive moat." After a day inside that interface, watching authorization flow from human to machine and back with zero friction, I understand why he calls the handoff the moat rather than the model.

Second, the certification building blocks are designed to be portable. SOI-3 was not a one-program milestone; the stated point is to bring foundational building blocks through the initial certification gates and then apply them across programs, the existing ones and the future ones. And the fleet strategy matches: the stated short-term goal is every aircraft in the fleet running the same technical capabilities and the same version of the software. The same brain on many airplanes. The system I flew behind is not a Caravan product. It is the product, wearing a Caravan.

Third, the catalyst is defined, by the company, on the record: when a new task order or expansion is announced, that is when the integration plan and the contract vehicle get laid out. Asked directly on our call whether the schedule was held up by Merlin, a subcontractor, aircraft availability, or SOCOM, the answer was flat: "No, I don't think there's a holdup." That denial is dated and on the record. The program sits post-CDR, in the vendor-and-partner assembly phase that precedes metal touching metal, under a $105 million ceiling IDIQ. The task order is the event this stock is waiting for, and when it comes, it arrives bundled with the plan.

One more macro point, because the demo made it concrete. Boeing's own industry outlook calls for roughly 660,000 new pilots worldwide over the next twenty years, against a training pipeline that everyone in the industry knows cannot produce them. The value proposition I watched at Quonset does not require replacing pilots to be worth money. Reduced crew workload, machine-disciplined fuel management, safety layers that never fatigue: these are line items on an operator's income statement long before any regulator changes a crew requirement. The pilot shortage is not a threat to this product. It is the demand curve.

What would make me wrong

I own this name, so the least I can do is state the other side plainly.

The structure is complex: a large convertible preferred stack with a 12 percent payment-in-kind dividend and a $5.00 conversion floor, which caps the death-spiral scenario but means the carry compounds quietly while the common waits. There is a lockup unlock in mid-September. Short interest at the last settlement sat near 90 percent of the float; the shorts and the index funds that arrived this summer cannot both be right about what happens next month. Reported revenue is going the wrong direction while G&A steps up. And the catalyst is a government schedule, which is to say, not a schedule. If the wait runs long, the tape will not be patient, and the preferred clock does not stop ticking while everyone waits.

Nothing I saw in Rhode Island refutes any of that. Hangars do not answer capital structure questions. What the visit answered is the question underneath all the others, the one that determines whether the rest is worth arguing about: is the product real, is it as far along as the public record claims, and are the people building it the kind who finish. The answers I brought home are yes, further than I expected, and I believe so.

The kicker

There is an old line in aerospace circles that, as flight control problems go, a cruise missile is just an airplane with somewhere to be. Engineers have trusted machines to fly themselves on one-way missions for half a century. What I watched at Quonset is the other half of that problem, the half that took fifty more years because it is harder: an aircraft that flies itself and asks permission, stays inside its boundaries, reads back its clearances in a voice with a sense of humor, and lands smoothly, gently, with investors in the back.

The submarines next door take a decade to build and the country pays what they cost, because some capabilities are not optional. Across the field sit the C-130s. In between, in a hangar with caution tape on the floor, a voice that is not a person is working the frequency.

Disclosure: Long MRLN, largest position in the book. This piece describes a company-hosted facility visit and demonstration flight. It contains no material non-public information. Conversations during the visit are characterized as my personal impressions only; no one at the company provided projections, financial information, timelines, or any information beyond the public record, and no statement in this piece should be read as made by or attributable to the company or its personnel except where drawn from published materials. All program facts, figures, and quotations come from public filings, company press releases, and the recorded management call previously published to subscribers. Nothing here is investment advice. Do your own work, and if you get the chance, go fly.

References

• Merlin, Inc. Form 10-Q for the quarter ended June 30, 2026, filed August 14, 2026, SEC EDGAR, CIK 0002028707

• Merlin, Inc. Form 8-K and Q2 earnings release, August 13, 2026

• Merlin press release, "Merlin Completes Autonomous Landing at EAA AirVenture Oshkosh," July 20, 2026

• Merlin press release, "Merlin Achieves Next Stage in Certification of Its Autonomous Flight Technology" (SOI-3, CAANZ), August 6, 2026

• Merlin press release, "Merlin Successfully Completes Critical Design Review for C-130J Autonomy Program with USSOCOM," June 4, 2026

• Merlin press release, Quonset State Airport flight test operations announcement

• USSOCOM C-130J autonomy IDIQ, $105M ceiling, FPDS / USAspending

• Shawarma Capital, "$MRLN: What Management Actually Said," August 21, 2026

• Textron Aviation, 3,000th Cessna Caravan family delivery announcement

• Boeing Pilot and Technician Outlook 2025 to 2044

• Nasdaq MRLN quote and short interest data, August 27, 2026

Not trimmed. Not rebalanced. Sold. Every position closed inside a few minutes, including the one that had been more than half my book for three weeks and was working.

The book is up 20.8% on invested capital in 4 months. That is the only performance number in this update and it is the one that matters, because everything underneath it just changed.

What Left

Five came off green, and those five carried most of the capital.

The two largest winners, IQE and Rubrik, both closed a shade under +138%. IQE I had held since the substrate work in the spring. Rubrik I had not touched in months, which is usually the sign a position is doing its job.

LPKF closed +55%. AXT closed +36.5% on a position that was three weeks old and had been up as much as +43% the week before.

The reds were Ouster at -3.2%, Coda at -13.7%, AmpliTech at -19.6%, Blue Moon at -27.9% and Amprius at -32.6%. I am not going to dress those up. Ouster I added to on August 12, six days before I sold it, which is the one I would take back. AmpliTech and Amprius were small sleeves that never got the second leg I underwrote. Blue Moon I wrote four parts on and I still think the asset is real, which is a separate question from whether I should have been holding it into this.

Agility through the CCXI trust closed at +18.9%. That name I had already traded once in early July and closed. I went back into it small after the August book update and closed it again on Tuesday with everything else. Two round trips, both green, neither large.

What Is Left

One position.

I have been publishing on Merlin since April. Thirteen parts. I have been wrong on pieces of it in public and corrected it in public. What changed Tuesday was not the thesis. The thesis has been the same since Part 1. What changed was the price against the thesis.

The stock printed $3.17 that morning. That was the low of the day and the lowest the name has traded since the de-SPAC. On the other side of that print sits roughly $183 million of cash, no debt, a company that cleared Stage of Involvement 3 with the New Zealand regulator, an aircraft that landed itself at Oshkosh in front of a crowd, and a sole-prime position on a $105 million IDIQ vehicle.

I did not want a 15% weight in that. I wanted the whole thing.

The reasoning is uncomfortable and I want to be direct about it. I had a book of ten names where nine of them were, at best, good. One of them was a situation. When you find a situation, the correct response is not to own a tenth of it alongside nine good ideas. Diversification is what you do when you do not know which one is the situation. I know which one it is.

That is either the best decision I have made or the one I will be writing about for a year. There is no version where it is neither.

The Tape Since

Three days is not a verdict. It is worth showing anyway because it is the only evidence that exists yet.

Merlin is up 6.3% from the exit print. Eight of the names I sold are lower than where I sold them, including AXT, which is down 13.6% in three sessions and had been the largest thing I owned. IQE gave back 15%. Ouster gave back 14%.

Both legs are working so far. Ask me in November.

What I Am Watching

The position is the whole book, so the watch list collapses into one list.

The task order. Remaining performance obligations at June 30 were $593 thousand against a $105 million ceiling. That gap is the entire near-term question. Ceiling is not revenue. What I need to see is obligated funding moving against the vehicle. Management told me directly that the task-order announcement is the event to watch and that it arrives bundled with the integration plan and the contract vehicle.

The board. The public job board has contracted from 30 requisitions to 23 since August 1, at the same time as intelligence that a new director-level hire is chartered to fill 41 to 43 technical roles. Those two facts do not sit together. A board that jumps from 23 toward 40 is the ramp signal. A board that keeps shrinking is something else.

September supply. There is a lock-up step in September and a conversion floor at $5.00. Below that floor the ratchet math gets worse for common holders. That is the single biggest reason this position could be right on the business and still hurt for another two quarters.

Cash. $183.5 million against roughly $25 million a quarter of operating burn is about seven quarters. The burn doubled year over year. The next raise happens either from strength or from the calendar, and which one it is tells you a great deal.

The Rules Have Not Changed

Everything I sold, I will keep covering. AXT, Blue Moon, IQE, LPKF, all of it. Coverage is not ownership and it never was. If one of them sets up again I will say so and I will say when I bought.

I will publish every mark on this position, up or down, on the same cadence I always have. A concentrated book is only defensible if it is a visible one.

Nothing here is advice. This is what I did with my own money and I would not recommend anybody copy the sizing.

The Merlin management full call tape lands separately this week. The AmpliTech post-mortem is already up.

This is the short version. Six things were said that change how I model the company, and one of them corrects something I have seen repeated everywhere, …

Read more

That is the disclosure. What follows is the autopsy, and it turned into something larger than a bad quarter.

Read more

Not investment advice. For educational purposes only. Do your own work.

Part 12 published on June 30 and ended with a promise. I wrote that Part 13 would mark the September lockup, the October preferred reset, and the second-quarter print as they resolved.

Six weeks later, here is the honest state of it. The company put more verifiable proof on the board in this window than in any comparable stretch since I started covering it. The stock printed an all-time low of $3.27 inside the same window. And on the morning of August 14 the 10-Q landed and told me that a piece of the capital structure I have been writing about for two months had already changed in May, in a way nobody covering this name reported.

I am going to do this chronologically, because the chronology is the argument. If you compress six weeks into a summary you lose the thing that actually matters, which is that every single piece of news in this window was either neutral or good, and the price went down anyway, and then the filing that arrived last told me the dilution was smaller than I had published.

Let me start with the tape, because it is the objection.

Read more

Read more

The phrase was "Foreign Ownership, Control or Influence."

That is a term of art. It comes from the Defense Counterintelligence and Security Agency, and it governs whether a company is permitted to hold a facility security clearance in the United States. A company under foreign ownership, control or influence cannot hold classified contracts without a mitigation instrument. The cleanest mitigation, the one that removes the question entirely rather than managing it, is to stop being foreign.

Amaero stopped being foreign. It completed a court approved scheme of arrangement, moved its ultimate parent from Australia to Delaware, and then, six weeks later, cleared the last two Australian directors off its board. The announcement says the reconstitution is "intended to address Foreign Ownership, Control or Influence considerations relevant to the Company's defense contracts."

Nine days before that, on 15 July, the same company confidentially submitted a draft registration statement on Form S-1 to the Securities and Exchange Commission.

Six days before that, on 1 July, it received its first production contract from Bechtel Plant Machinery Incorporated, which is the prime contractor for the United States Naval Nuclear Propulsion Program.

None of these three events was covered by any outlet I could find in English beyond a wire pickup of the press release. The company has four documents in the entire SEC EDGAR full text index, and all four of them are somebody else's filings.

This is the first of a series on Amaero. It trades in Australia as 3DA and in the United States as AMROF. I do not have a position. By the end of this piece you will understand exactly what it makes, why the Navy cares, what the Velo3D relationship is actually worth as opposed to what the headline says, and the single number in its most recent filing that makes the whole thing urgent rather than merely interesting.

That number is 2.28.

What this company physically does

Start with the object, because the object is unusual.

Amaero makes metal powder. Not powder in the sense of a fine dust you might imagine, but spherical particles of tightly controlled diameter, produced by melting a metal electrode and blowing the resulting stream apart with inert gas so that surface tension pulls each droplet into a sphere before it freezes. The process is called Electrode Induction Melting Inert Gas atomization, universally shortened to EIGA. The sphericity matters because these powders are spread in layers forty microns thick across a build plate and then selectively melted by a laser. A powder that does not flow does not spread. A powder that does not spread does not print.

The company runs three EIGA Premium atomizers at a facility at 130 Innovation Drive in McDonald, Tennessee, which is a small town outside Chattanooga. One of those atomizers is dedicated to refractory alloys. Two are dedicated to titanium. Entering fiscal 2027 that installed base represents roughly 200 metric tonnes per year of refractory alloy powder and roughly 480 metric tonnes per year of titanium alloy powder.

The titanium is the volume business. The refractory is the business that explains everything else.

Refractory metals are the metals that do not melt when everything else has. Niobium melts at 2,477 degrees Celsius. Tantalum at 3,017. Tungsten at 3,422. For comparison, the nickel superalloys that make up the hot section of a jet engine start losing structural integrity somewhere north of 1,000 degrees and are kept alive by film cooling and thermal barrier coatings. When you need a part that holds its shape in a gas stream hotter than that, with no cooling available and no time to design a cooling circuit, you reach for refractory metal.

The specific alloy that matters here is C103. Its composition is niobium with roughly ten percent hafnium and one percent titanium, which is why you will see it written as Nb-10Hf-1Ti. It was developed in the 1960s. It flew on the Apollo service module propulsion system. It has been used in rocket engine nozzle extensions for sixty years, because a nozzle extension is the canonical application: the gas is extremely hot, the part is thin, the part is large, and there is no coolant available at the exit plane because the coolant has already been used upstream.

C103 has one significant weakness. It oxidizes catastrophically in air at temperature. Bare C103 exposed to atmosphere above about 400 degrees Celsius will not merely tarnish, it will consume itself. This is why C103 parts are coated, historically with a silicide coating designated R512E. In a vacuum, which is where rocket nozzle extensions operate, the problem does not arise. In atmospheric flight, which is where hypersonic vehicles operate, it very much does, and the coating becomes a life limiting component rather than a finish.

That distinction between vacuum applications and atmospheric applications is the entire hypersonics materials debate compressed into one sentence, and I will come back to it, because it bears directly on how much of the demand narrative around this company is real.

The second thing Amaero does is called PM-HIP, which stands for Powder Metallurgy Hot Isostatic Pressing. You take metal powder, seal it in a shaped can, and subject it to high pressure and high temperature simultaneously until the powder consolidates into fully dense metal in the shape of the can. What comes out is a near net shape part with, in the company's phrasing, "forged-equivalent material properties and microstructure."

That phrase is doing enormous work, and I want to be precise about why.

A large forging requires a large forging press and a supply chain of open die and closed die forge shops. The United States has a well documented shortage of that capacity, particularly for large components in nuclear and naval applications. The forging base contracted for thirty years and the qualification burden for a new forge supplier is measured in years. PM-HIP is not a cheaper forging. It is a different route to a component with equivalent properties that does not require a forging press at all. If you believe the qualification data, it substitutes capital equipment that does not exist for capital equipment that does.

The Navy has said it believes the qualification data. According to the company's July investor presentation, the Navy "recently announced support for PM-HIP as a mature manufacturing technology and a viable alternative for large casting and forging supply chain challenges." That sentence is a company characterization of a Navy position and I have not independently sighted the Navy statement, so treat it accordingly. But the contract flow, which I will come to, is consistent with it.

How it got here, which is not the story it tells

The company's investor materials present a clean arc: build capacity, qualify powders, win defense contracts, move to America. The actual history is messier and more informative, and I think it matters for judging management.

Amaero began at Monash University in Melbourne. Monash established the Monash Centre for Additive Manufacturing in 2010. In March 2013, Professor Xinhua Wu and other MCAM staff created Amaero to commercialise what the centre was producing. Amaero International Ltd was incorporated on 21 May 2019 purely as a holding vehicle for a listing, and it floated on the ASX on 5 December 2019, raising A$8.0 million by issuing 40 million shares at twenty cents.

What listed in 2019 was not a powder company. The ASX admission describes a laser additive manufacturing bureau. Amaero printed parts for customers. It did not make the feedstock.

Monash granted two intellectual property licences dated 18 October 2019, one exclusive and global covering titanium, one non exclusive covering scandium, carrying a minimum annual fee of A$50,000 from FY2023 and a A$150,000 buyout option. That university connection is now effectively gone. The FY2025 annual report records no payments to Monash and carries no licence or patent intangibles. Monash Investment Holdings held exactly 18,315,178 shares from the IPO through FY2023, diluting passively from 11.11 percent to 4.39 percent without ever buying more, and exited the register entirely during FY2024. The original 2013 operating entity, Amaero Engineering Pty Ltd, had its Australian Business Number cancelled on 7 July 2026.

The pivot from bureau to powder was attempted first in Australia, and it failed. In July 2021 the company announced it would build a world class titanium powder plant in Australia. It revised the associated revenue guidance in September 2021, secured a lease in January 2022, and issued an update in July 2022. That plant is not what exists today. What exists today is in Tennessee.

The turn came in September 2022, when Hank Holland, managing partner of the US private equity firm Pegasus Growth Capital and then a non executive director, became Executive Chairman. The founding era chairman stepped back to a non executive seat. From that point the company reoriented around a United States manufacturing base, United States customers, and eventually a United States domicile.

I think that history should be read two ways at once, and most coverage picks one.

The unflattering reading is that this is a company that has changed its business model twice, abandoned a announced manufacturing plan in one country, and is now on its third strategy in seven years, with a share count that went from 175 million to 953 million funding the journey.

The flattering reading is that the second pivot worked. Revenue went from under half a million dollars to A$18.1 million. The capital program completed on budget. The customer list now includes a naval nuclear prime and the Department of War. A management team that abandoned a plan that was not working and executed a better one is doing its job, and the fact that the current strategy is the third one is only damning if you believe the current one is also wrong.

What I take from it is narrower and more useful: this is a private equity controlled company executing a private equity playbook. Acquire control of a listed shell with useful technology, redirect it at a defensible end market, fund it through repeated raises you lead yourself, build the asset, then exit into a deeper capital market at a higher multiple. That is not a criticism. It is a description, and it explains the redomiciliation and the S-1 far better than any of the strategic language in the presentation does.

Why this is hard to make

It is worth spending a moment on why a company can build a defensible position around what is, on its face, a commodity. Powder is powder. Except that it is not, and the reasons are physical rather than commercial.

Consider what you are asking the process to do with C103. You need to melt a metal at roughly 2,477 degrees Celsius, hold it liquid long enough to break the stream into droplets, and freeze those droplets into spheres, all without letting the melt touch anything that would contaminate it and without letting it see oxygen.

Every one of those constraints is severe at that temperature.

Take contamination first. The conventional way to melt metal for atomization is in a crucible, typically ceramic. At 2,500 degrees, niobium is aggressively reactive and will attack essentially any ceramic you put it in, dissolving the crucible into the melt. The resulting powder carries ceramic inclusions, and an inclusion in a printed part is a crack initiation site. In a rocket nozzle or a submarine component, a crack initiation site is the whole ballgame.

This is why the process Amaero uses is EIGA rather than conventional gas atomization. Electrode Induction Melting Inert Gas atomization is crucible free. A pre alloyed bar of the finished chemistry is fed downward through an induction coil, which melts only the tip. The molten metal drips directly into the gas nozzle. Nothing but the alloy itself ever touches the melt. That single design choice is what makes reactive and refractory alloys atomizable at all, and it is why you see EIGA specified for titanium and refractory work while conventional crucible atomization dominates for steels and nickel alloys.

Take oxygen second. Titanium and niobium both dissolve oxygen interstitially, meaning oxygen atoms sit in the gaps of the crystal lattice rather than forming a discrete oxide. Interstitial oxygen makes the metal harder and much more brittle. Aerospace specifications for titanium powder therefore carry hard oxygen ceilings, often in the range of 1,300 to 1,500 parts per million for Ti-6Al-4V, and every time powder is handled, sieved, or reused it picks up more. A producer's ability to hold oxygen low through melting, atomization, classification and packaging is a genuine process capability and it is not easily copied.

This is also why the Auburn qualification detail I mentioned earlier matters more than a casual reading suggests. ASTM F3635 is the specification covering additively manufactured C103. Meeting Class B after a 2200 degree Fahrenheit heat treatment means the material was assessed as built and heat treated, which is the condition a part actually flies in, rather than as loose powder in a bottle. AMS7015 and ASTM F3001 for the titanium are similarly the printed material specifications rather than feedstock specifications.

Take yield third, because yield is where the economics live. Gas atomization does not produce one particle size. It produces a distribution, and only a fraction of that distribution falls inside the window a laser powder bed fusion machine can use, typically something like 15 to 45 microns. Everything coarser has to be sold into other processes at lower value, recycled, or scrapped. When your feedstock is niobium and hafnium, the difference between a process that yields 40 percent into the useful cut and one that yields 60 percent is not a rounding error, it is the gross margin.

Amaero's own boilerplate makes exactly this claim. It says the company "has commissioned advanced gas atomization technology with an industry leading yield of AM powder." I cannot verify that claim independently and it is precisely the sort of assertion a company makes about itself, so I am recording it as a claim rather than a fact. But note that it is the correct thing to claim. A management team that understood this business would compete on yield, and a management team that did not would compete on price.

Take scale last. Three atomizers producing 680 tonnes a year sounds modest next to a steel powder plant, and it is. But refractory powder is not sold by the tonne at commodity prices. It is sold by the kilogram at prices set by the input metal and the difficulty of the process. The relevant comparison is not tonnage against a steel producer, it is whether 200 tonnes a year of refractory capacity is large or small against the total Western demand for refractory AM powder, and the honest answer is that nobody publishes that number reliably. What can be said is that 200 tonnes per year of C103 at the hafnium content described later in this piece would consume a meaningful fraction of annual world hafnium production, which tells you the capacity is not small relative to the input chain even if it looks small in absolute terms.

The Velo3D relationship, and what it is actually worth

Because the question that prompted this research was about the Velo3D supply agreement, let me deal with it directly and then explain why I think it is the least interesting thing about Amaero.

Both companies announced it on 29 April 2025. Neither has ever disclosed an execution date or an effective date. It surfaced in Velo3D's SEC filings two weeks later, on 13 May 2025, inside the supplementary slides attached to its first quarter 8-K, under a page headed "Velo3D and Amaero forge strategic alliance to propel metal 3D printing innovation."

The counterparty named in Velo3D's filing is "Amaero Advanced Materials and Manufacturing, Inc.", which is the US operating subsidiary rather than the listed parent.

The scope of exclusivity is wider than most summaries report, and the asymmetry in it is the whole point. The identical wording in both releases is that Amaero "will serve as the exclusive supplier to Velo3D for Niobium C103 and other refractory alloy powders, including Molybdenum, Tantalum, Tungsten, and Zirconium alloys" and "will be preferred supplier to Velo3D for titanium alloy powders."

So Amaero is exclusive across five refractory alloy families, not just C103, and preferred on titanium. Velo3D's side of the bargain is operational rather than financial: qualify Amaero's powders, develop proprietary print parameters exclusively for Amaero's C103 and refractory alloys across the entire Sapphire printer family, and supply those parameters with new machine sales.

Critically, the exclusivity runs one way. It binds Velo3D's purchasing. It does not bind Amaero's selling. Amaero remained free to sign parallel agreements with other counterparties and has done exactly that, with Titomic, Knust-Godwin, ADDMAN and United Performance Metals.

The only disclosed termination right belongs to Amaero. Its release states it "reserves the right to terminate the agreement if qualification of C103 and Ti64 powders is not achieved by 30 November 2025." Velo3D's release of the same day does not mention the deadline or the right. Amaero declared the condition satisfied on 8 July 2025, one hundred and forty five days early, on the strength of the Auburn testing.

Now the headline number, which deserves more scrutiny than it usually gets.

Amaero said A$35 million, "or US$22.6 Million based on the exchange rate on April 25, 2025." That implies an AUD/USD rate of 0.6457. Velo3D said "approximately $22 million" and footnoted it as "calculated based on an USD/AUS exchange rate of 0.6380 on 25 April 2025." Two counterparties, the same Australian dollar figure, the same cited date, two different exchange rates and a US$600,000 gap between the translations.

The cited date is 25 April 2025, which was Anzac Day. The Reserve Bank of Australia published no exchange rate that day.

This is a small thing and I am not suggesting anything improper. It is the kind of discrepancy that arises when two investor relations teams convert a number independently against different dealer sources. I raise it because it is a useful reminder of what this figure is: a marketing translation of an estimate, not a contracted amount.

Then on 6 August 2025, Velo3D filed again, and this is the substantive one. Auburn University's National Center for Additive Manufacturing Excellence completed qualification of Amaero's powders. The results as stated: C103 niobium met ASTM F3635 Class B after a 2200 degree Fahrenheit heat treatment, and Ti-6Al-4V met AMS7015 and ASTM F3001. Over 1,000 kilograms of powder was to be delivered to Velo3D's production floor in the third quarter of 2025.

That is a real qualification at a real independent laboratory against real published standards. It is not a memorandum of understanding. Powder that meets F3635 Class B is powder a customer can build flight hardware from without writing its own material specification first, and the heat treatment detail tells you the qualification was done on printed and post processed material rather than on the powder in the bottle.

Here is where I want to be careful, because this is where most coverage of this relationship goes wrong.

The A$35 million is not contracted revenue. Read the construction of the disclosure: the value is derived from Velo3D's own demand estimates over the five year term. There is no minimum volume commitment disclosed, no take or pay provision disclosed, and no schedule disclosed. Amaero is the exclusive supplier of C103 and refractory alloys to Velo3D, and the preferred supplier of titanium. Exclusivity of that shape is an obligation on Velo3D to buy from Amaero if it buys at all. It is not an obligation to buy.

So the correct way to hold this number is as a ceiling on a five year opportunity conditional on Velo3D's own volumes, not as a floor under Amaero's revenue.

And there is a harder confirmation of that, which I think is the single most important fact about this relationship and which I have not seen reported anywhere.

Velo3D never treated the agreement as a material definitive contract.

There is no Item 1.01 current report. There is no Exhibit 10 filing of the agreement itself. When a public company enters a material definitive agreement outside the ordinary course of business, Item 1.01 of Form 8-K requires disclosure within four business days and the contract is normally filed as an exhibit. Velo3D did neither.

More telling still: Velo3D's Form 10-K for fiscal 2025, filed 31 March 2026, contains zero occurrences of "Amaero", zero of "C103", zero of "refractory" and zero of "niobium". The five year exclusive supply agreement its own press release called a strategic alliance does not appear in its annual report.

And the audited commitments note rules out any minimum obligation affirmatively. The FY2025 10-K states that purchase obligations with suppliers "are cancellable in whole or in part prior to shipment," discloses non-cancellable purchase commitments of US$7.7 million for parts and assemblies, and then says the company "has no other commitments and contingencies, except for the operating leases." The Q1 2026 10-Q carries the same structure with the parts and assemblies figure at US$24.4 million.

There is no take or pay. There is no minimum volume. The customer's auditors have confirmed it in writing.

I want to be fair about what this does and does not mean. It does not mean the relationship is fake. The qualification happened, the powder shipped, and the exclusivity on refractory alloys is real and documented by both sides. What it means is that the A$35 million figure has no contractual force whatsoever, and that the counterparty's own accountants do not regard the arrangement as creating an obligation worth disclosing. Anyone carrying A$35 million of Velo3D revenue in a model for this company is carrying a number that exists in two press releases and nowhere else.

Which raises the counterparty question, and the counterparty question is not comfortable.

Velo3D's revenue went from US$77.4 million in 2023 to US$41.0 million in 2024. That is not a growth company having a soft year, that is revenue nearly halving. Gross margin was negative 33.9 percent in 2023, negative 5.1 percent in 2024, and negative 11.7 percent in the second quarter of 2025. The company has never, in the periods disclosed in these filings, sold a printer for more than it cost to build. Its own 8-K risk factor list includes, in order, "the company's ability to continue as a going concern" and "the company's ability to raise additional capital in the near-term."

There is genuine good news on this front and I do not want to bury it. Velo3D has moved from the over the counter market back onto Nasdaq. SEC records now list its tickers as VELO and VLDXW against exchanges Nasdaq and OTC. Anyone reading a source that still calls it VLDX is reading something stale. It guided in August 2025 to US$50 million to US$60 million of 2025 revenue, gross margin above 30 percent exiting 2025, and EBITDA profitability in the first half of 2026. If it hit those marks it is a materially different company than the one in the table above.

But note what Velo3D's other 2025 deals look like. A US$15 million master services agreement with Momentus, paid in Momentus common and convertible preferred stock rather than cash, with Velo3D's stake capped at 9.99 percent of Momentus. A US$4 million agreement with Vaya Space. These are the deal structures of a company managing its cash position carefully.

My conclusion on Velo3D as it relates to Amaero: it is a legitimate, Nasdaq listed, technically credible customer whose printers are genuinely well suited to the alloys Amaero makes, and the Auburn qualification is a real asset that outlives the commercial relationship because it is a qualification of the powder, not of the pairing. But the A$35 million should be modelled as optionality. If you are buying Amaero for the Velo3D contract, you have bought the wrong thesis.

The right thesis is on the other side of the building.

The submarine

On 1 July 2026, Amaero announced a low rate initial production contract from Bechtel Plant Machinery Incorporated for piping components supporting the submarine industrial base. The value was US$344,000.

Three hundred and forty four thousand dollars. Against a company with a market capitalization in the low hundreds of millions. On its face this is a rounding error and it would be entirely reasonable to skip past it.

Do not skip past it. Read the next sentence in the quarterly report:

"Though not announced, a total of six contracts have previously been awarded by BPMI. Between the previous work and this latest contract, work ranges from development and demonstration to first article and production."

Six prior contracts, never announced. A company that issues an ASX release for a webinar recording did not announce six contracts from a naval nuclear prime. I can think of benign reasons, principally that each was individually immaterial under the continuous disclosure threshold, and that is likely the correct explanation. But the aggregate picture it reveals is a two year qualification relationship that the market has never seen the shape of.

Understand what BPMI is. Bechtel Plant Machinery Incorporated is the prime contractor to the Naval Nuclear Propulsion Program, the joint Navy and Department of Energy organization that designs, builds and maintains the reactors in United States submarines and aircraft carriers. Naval Reactors is the most conservative engineering organization in the American government, and by reputation possibly the most conservative in the world. It is the direct institutional descendant of Rickover. Material qualification inside that program is measured in years and the program does not qualify suppliers speculatively.

The progression Amaero describes is the standard defense qualification ladder: development, then demonstration, then first article, then low rate initial production, then full rate production. Amaero has just stepped onto the fourth rung. The company's own framing is that this marks "Amaero's transition from PM-HIP validation to production."

And then the number that actually matters, from the July investor presentation:

"BPMI has previously announced estimated demand for up to 400 PM-HIP components per year to support current and near-term shipbuilding needs."

Four hundred components a year against a first production award of US$344,000. I do not know the unit economics of a submarine piping component and I am not going to invent them. But the ratio between the demonstrated award and the stated addressable demand is the largest single asymmetry in this company, and unlike the Velo3D number it is anchored on a customer's published requirement rather than a customer's demand forecast used for a press release.

This also explains the December 2025 Letter of Support from the Department of the Navy, which the company describes as having been "solidified" by the LRIP award, and it explains something else that was otherwise slightly odd. It explains why a company with a market capitalization this small bothered to spend real money redomiciling itself to Delaware.

Why a company moves countries

Between February and July 2026, Amaero executed the following sequence.

February 2026, announced its intention to redomicile to the United States. March 2026, nominated Tim Johnson to the board. May 2026, distributed the scheme booklet. 5 June 2026, obtained shareholder approval at an extraordinary general meeting. 11 June 2026, the schemes became effective and the new Delaware entity was admitted to the ASX official list. 22 June 2026, redomiciliation became effective. July 2026, completed a PCAOB audit and confidentially submitted the Form S-1.

The mechanism was a Scheme Implementation Deed, executed under section 411 of the Australian Corporations Act, and there were two schemes, one for shares and one for options. Amaero Inc., a Delaware corporation, became the ultimate parent. Existing shareholders received CHESS Depositary Interests, which continue to trade on the ASX under the same 3DA ticker.

The conversion ratio is the detail that trips people up, so hold onto it: shareholders received 40 CDIs per one share of Amaero Inc. common stock. The ASX line is priced in fortieths. When this company eventually lists in the United States, the share price Americans see will be roughly forty times the ASX quote, which converts a stock that looks like a twenty four cent penny share into something in the range of seven dollars. That is not a trivial cosmetic point. It is the difference between a security most American institutions cannot touch and one they can.

Now, the reasons the company gives for the move, taken from the July presentation, are these: alignment with United States defense and sovereign manufacturing growth, greater visibility with US customers, improved comparability against US listed peers, access to a larger investor base, potential for improved valuation and liquidity, enhanced access to lower cost debt and equity, simplification for mergers and partnerships, and positioning for a potential US listing.

All of that is the language every company uses when it redomiciles, and taken alone it would be unremarkable boilerplate.

But one bullet is not boilerplate. The presentation says the move "positions Amaero to satisfy U.S. foreign ownership, control and influence (FOCI) requirements, supporting eligibility for classified defense contracts."

And the quarterly report says the redomiciliation strengthens "its ability to pursue classified U.S. Department of War and federal contracts by mitigating foreign ownership concerns."

Classified. Not sensitive, not controlled, not ITAR restricted. Classified.

A company does not restructure its parent entity, spend A$2.8 million in a single quarter on redomiciliation and IPO costs, and clear its foreign directors off the board in order to bid work it is already eligible to bid. The FOCI language, repeated across three separate disclosures in three separate documents, is the company telling you that there is work it wants which it cannot currently hold, and that it has now removed the obstacle.

The board reconstitution on 4 August completes it. Jamie Levy and Alistair Cray, both non executive directors, resigned effective that day. Both were retained on consultancy agreements through 31 December 2026 to handle Australian investor relations and corporate advisory. That structure tells you the exits are cooperative rather than acrimonious, which is worth saying plainly, because two directors leaving on one day would otherwise be a governance flag. The board is now, in the company's words, "a fully US board."

I want to flag the thing I cannot verify, because it is the load bearing uncertainty in this section. I do not have evidence that Amaero holds or has applied for a facility security clearance. FOCI mitigation is a prerequisite for one, not a substitute. The company has told us it has removed a barrier. It has not told us it has walked through the door, and I am not going to assume it has.

The S-1

On 15 July 2026 Amaero announced it had confidentially submitted a draft registration statement on Form S-1 relating to a proposed initial public offering of shares of its common stock in the United States. Share count and price range are not yet determined. The offering is subject to market conditions and completion of SEC review. The announcement was made under Rule 135, which is the safe harbor that lets an issuer say a registration is coming without the statement itself constituting an offer.

Two things are notable about the way this was done.

First, the company completed a PCAOB audit of its historical US GAAP consolidated financial statements for calendar years 2024 and 2025, and those audited statements went into the submission. That is not a small undertaking for a company this size and it is a genuine credibility marker. A PCAOB audit of US GAAP accounts is the thing that separates companies that intend to list from companies that talk about listing.

Second, the target window. The July presentation says a potential US listing is "targeted for late CY2026 / early CY2027, subject to market conditions and the completion of the SEC review process."

That is roughly four to eight months from today.

So here is the shape of the setup. A company completes its redomiciliation, completes a PCAOB audit, files confidentially, tells you it is aiming at a listing inside twelve months, and continues to trade in the meantime on the Australian Securities Exchange at a price set by Australian retail flow and on the American over the counter market at whatever the spread allows.

I am not going to tell you that pre listing accumulation is a free lunch, because it is not, and I will get to the reasons in the section on what breaks this. But the structural observation is straightforward: the buyer base that is most likely to reprice this security is currently unable to buy it in any size.

The contract Amaero got from the Department of War

On 22 July 2026, one day after the quarterly, Amaero announced a contract valued at approximately A$6.5 million, or US$4.5 million, from the United States Department of War for the development of alternative refractory alloy powders.

The details as disclosed: a thirteen month program expected to complete in August 2027, aimed at "reducing the cost of refractory alloy powders while reducing reliance on foreign sources for critical raw materials," with progress payments based on project milestones.

Three observations.

The milestone payment structure matters more than it sounds. A development contract with progress payments is working capital positive, or at worst neutral. For a company in Amaero's cash position, and we are about to look hard at that cash position, the difference between a milestone paid program and a cost incurred then invoiced program is the difference between a contract that helps and a contract that hurts.

The stated objective is the interesting part. "Reducing reliance on foreign sources for critical raw materials" is not a description of the powder business, it is a description of the input supply chain. C103 is niobium plus hafnium. Niobium production is overwhelmingly concentrated in Brazil, where CBMM alone accounts for the substantial majority of global supply. Hafnium is scarcer and stranger: it is produced almost entirely as a by product of separating hafnium out of zirconium to make nuclear grade zirconium alloys, which means global hafnium supply is a function of nuclear fuel cladding demand and not of hafnium demand. Annual world production is measured in tens of tonnes.

If the Department of War is paying Amaero to develop alternative refractory alloy powders that reduce foreign input dependence, the most natural reading is that somebody in the department has looked at the C103 supply chain and concluded that hafnium is a chokepoint. That is a real strategic tailwind for a domestic producer. It is also, and this is the part a bull case tends to skip, an acknowledgement that the alloy Amaero is currently known for has an input problem that the government would like to engineer around.

Both of those things are true simultaneously. A programme to develop alternatives to C103 is good news for the company doing the developing and ambiguous news for the long term volume of C103 itself.

The third observation is scale. A$6.5 million against FY2026 revenue of A$18.1 million is a large single contract for this company. It is 36 percent of a year of revenue, spread across thirteen months, and it is development work rather than production, which typically carries different margin characteristics than powder sales.

The hafnium problem

I want to take the Department of War contract's stated purpose seriously, because when a government agency pays a company to reduce reliance on foreign raw materials, the interesting question is which material it means.

It means hafnium. Here is why.

C103 is nominally ten percent hafnium by weight. Hafnium is not mined. There is no hafnium mine anywhere in the world and there never has been. Hafnium occurs in zircon, the zirconium silicate mineral, at a ratio of roughly fifty parts zirconium to one part hafnium, and the two elements are chemically almost identical, which is what makes them difficult to separate and also what makes the separation necessary.

The necessity is nuclear. Zirconium has an extremely low neutron absorption cross section, which makes it the ideal material for fuel cladding in a nuclear reactor. Hafnium has an extremely high neutron absorption cross section, which makes it an excellent control rod material and a catastrophic contaminant in cladding. So the nuclear industry must separate them, and it does, at considerable expense, producing nuclear grade zirconium as the product and hafnium as the by product.

That single fact governs the entire hafnium market. World hafnium supply is set by demand for nuclear fuel cladding, not by demand for hafnium. If every hafnium consumer on earth doubled their orders tomorrow, supply would not respond, because nobody separates zircon in order to get hafnium.

The numbers, drawn from industry and trade estimates rather than a single authoritative source, so treat them as directional: total refined world hafnium production runs at roughly 70 to 75 tonnes per year. Production is concentrated in four countries, being France, the United States, China and Russia. Framatome in France is estimated at around 43 percent of world output, roughly 30 tonnes a year, which makes a single French company the largest hafnium producer on the planet.

Read that list again in the context of a Department of War contract about reducing reliance on foreign sources. The largest producer is French, which is to say allied but foreign. Two of the remaining three are China and Russia. The United States is one producer among four in a market of about 75 tonnes.

The price has behaved the way you would expect a chokepoint to behave. Hafnium has risen roughly eightfold since 2020. It reached the range of US$6,300 to US$7,000 per kilogram in Rotterdam by late 2025 and has been quoted above US$12,000 per kilogram on some dealer benchmarks in early 2026. One reference series has it moving from US$4,364.80 per kilogram in January 2025 to US$9,499.90 in January 2026, a gain of about 118 percent in a year.

Now do the arithmetic that the company does not do for you. At ten percent hafnium content, one kilogram of C103 contains one hundred grams of hafnium. At US$9,500 per kilogram, that is US$950 of hafnium in every kilogram of alloy. At the US$12,000 dealer prints, it is US$1,200. Before you have paid for niobium, before you have melted anything, before you have run a single kilowatt through an induction coil, and before anybody has earned a margin.

That is the cost structure of C103, and it is the reason the Department of War is funding alternatives.

Projected demand adds the second half of the picture. Estimates put hafnium demand near 180 to 190 tonnes by 2030 against a supply base struggling to exceed 75 tonnes. The 2023 gap was already estimated at 44 to 49 tonnes. If those projections are anywhere near right, the gap by 2030 exceeds 105 tonnes, which is more than current world production.

For Amaero this cuts three ways and an honest analysis has to hold all three at once.

It is bullish in the sense that a company already qualified on a hafnium bearing alloy, in a market where the input is scarce and Western supply is thin, sits on the right side of a structural shortage. Scarcity of input is a barrier to entry for everyone, including the entrants.

It is bearish in the sense that Amaero has not disclosed secured hafnium supply. A converter buying a scarce input at spot and selling a qualified output on multi year agreements is exposed on both sides of the trade. If hafnium doubles again and Amaero's supply agreements do not have pass through provisions, the margin on C103 goes to the input, not to the shareholder. I have seen no disclosure on this either way, which is itself the finding.

And it is ambiguous in the sense that the Department of War is paying to engineer around the problem. If that thirteen month programme succeeds in delivering an alternative refractory alloy powder with less hafnium or none, then the long term volume of C103 specifically may be lower than the hypersonics narrative implies, even as Amaero is the company being paid to build the replacement. Being paid to obsolete your own product is a better position than having somebody else paid to do it, but it is not the same as unbounded C103 demand.

The single most valuable disclosure Amaero could make, and the one I will be looking for in the Form S-1, is the structure of its raw material supply. An S-1 risk factor section has to address input concentration. An ASX announcement does not.

The forging base

The other half of the business needs the same treatment, because the PM-HIP opportunity rests on a supply chain failure most investors have never had reason to learn about.

The United States lost its heavy forging capacity gradually and then structurally. Large open die and closed die forging requires enormous presses, and the number of presses in the Western hemisphere capable of the largest naval and nuclear components is small enough to count. Those presses are old, they are heavily booked, and the qualification burden to add a new supplier into a naval nuclear application runs to years. The consequence is lead times measured in years for components that a shipbuilding programme needs on a schedule.

This is not a marginal issue for the Navy. It is one of the primary constraints on submarine production rate, and submarine production rate is currently a matter of national policy attention given the Virginia class build rate, the Columbia class ballistic missile submarine programme, and the commitments made under the trilateral security arrangement with Australia and the United Kingdom.

PM-HIP addresses it from a different direction. Rather than deforming a large billet under a press, you consolidate powder in a shaped container under isostatic pressure and temperature. The result is a near net shape component with a fine, uniform, isotropic microstructure and no directional grain flow, which for many applications is equivalent to or better than a forging. Crucially, it does not need a forging press. It needs a hot isostatic press, powder, and a qualified process.

The reason this has not simply replaced forging everywhere is qualification. Naval nuclear components are governed by material specifications written decades ago that assume a wrought product form. Getting a PM-HIP component accepted means demonstrating equivalence across mechanical properties, fracture toughness, weldability and inspection response, and doing it to the satisfaction of an organisation with an institutional memory that treats novelty as risk.

Which is what makes the July disclosures significant. The Navy has, according to the company and corroborated by trade press coverage, recognised PM-HIP as a mature and well established process that meets rigorous technical requirements and compresses production timelines, and has described it as a viable alternative for large casting and forging supply chain challenges. The Maritime Industrial Base Program is described as facing long lead times and capacity constraints in exactly that supply chain.

So the sequence reads: the Navy has a forging bottleneck, the Navy has blessed PM-HIP as a route around it, BPMI has published a requirement for up to 400 PM-HIP components a year, and Amaero has just moved from first article to low rate initial production after seven contracts and two years of work.

The gap between US$344,000 of production revenue and 400 components a year is the entire investment case for this half of the business. I want to be careful not to let the size of the addressable number do the work of an argument. Amaero is not the only company capable of PM-HIP, the 400 component requirement is not exclusively addressable by any one supplier, and low rate initial production is a rung on a ladder rather than a destination. But the direction of travel is documented in primary sources, and the customer is the least promotional organisation in American industry.

The numbers

Now the financial position, which I have taken entirely from the Appendix 4C and the quarterly activities report filed on 22 July 2026, and from the FY2025 annual report filed 18 September 2025. Amaero's fiscal year ends 30 June.

Revenue first.

FY2024 revenue was A$463,856. FY2025 was A$3.8 million. FY2026 was A$18.1 million. That is a 376 percent increase on FY2025, and it landed inside the A$18 million to A$20 million guidance the company gave in January 2026, which is worth noting because guidance that is met is a data point about management quality independent of the absolute number.

The quarterly path inside FY2026 is not a smooth ramp and the shape tells you something.

Q1 A$4.7 million, Q2 A$3.1 million, Q3 A$2.6 million, Q4 A$7.8 million. Revenue went down for two consecutive quarters in the middle of a year that finished up 376 percent. That is what a business looks like when individual orders are large relative to total revenue and shipment timing dominates. It is also a warning about extrapolating any single quarter in this company.

Q4 would have been A$9.1 million and the full year A$19.5 million except that A$1.3 million of contracted titanium orders were pushed into Q1 FY2027 by a production pause I will describe shortly.

The trailing twelve month view smooths it out.

A$7.4 million, A$9.9 million, A$11.8 million, A$18.1 million. Year on year TTM growth at each of those points was 313, 384, 347 and 376 percent.

FY2026 revenue splits into A$15.6 million of powder and A$2.6 million of PM-HIP. PM-HIP is fourteen percent of revenue and it is the segment attached to the submarine industrial base, which means the strategically most important business line is currently the smaller one. In Q4 specifically, growth was supported by shipments of 30 PM-HIP components plus titanium and refractory powders across 19 unique customers.

Backlog next, and note the date carefully because the company gives two.

Backlog at 30 June 2026 was A$14.6 million. Between then and 22 July the company added approximately A$8.5 million, which includes the A$6.5 million Department of War contract and A$1.8 million of refractory powder work. Total backlog as of 22 July was A$23.1 million, of which A$12.9 million relates to work currently scheduled for completion by 31 December 2026. Backlog spans 26 unique customers. The company also discloses 28 or more tonnes of titanium powder contracted for the third calendar quarter of 2026.

A$23.1 million of backlog against A$18.1 million of trailing revenue is a healthy ratio for a company at this stage. The A$12.9 million scheduled by year end is the number to hold management against at the December quarterly.

Then capacity, which is the part of the story where the capital has already been spent.

Capacity went from 200 tonnes per year in June 2024 to 440 in June 2025 to 680 in June 2026, with 920 guided for June 2027 once the fourth EIGA atomizer commissions. Capital deployed tracked it: A$18.2 million, A$51.6 million, A$80.0 million, and A$94.0 million guided.

The company completed its original three year A$72 million capital investment program in the June quarter, on schedule and on budget, with the commissioning of the third EIGA atomizer ahead of schedule. The next phase is the fourth atomizer plus an argon gas recycling plant, the latter expected to commission in early calendar 2027. Argon recycling is an operating cost item rather than a capacity item and it matters because inert gas consumption is one of the larger variable costs in gas atomization.

Completing a multi year capital program on budget is genuinely uncommon in small cap industrials and management deserves credit for it. It also means the story from here is no longer about building capacity. It is about filling it. At A$18.1 million of revenue against 680 tonnes of installed capacity, utilization is low. Whether that gap closes is the entire question.

The number that makes this urgent

Here is the Appendix 4C.

Liquidity at 30 June 2026 against the quarter's cash consumption.

Cash and cash equivalents at 30 June 2026: A$19.43 million. Down from A$33.42 million at 31 March. Net cash used in operating activities in the quarter: A$10.90 million. For the full year: A$35.28 million. Net cash used in investing activities in the quarter: A$9.38 million. For the full year: A$36.44 million. Net cash from financing for the year: A$74.61 million.

Total financing facilities A$37.98 million, drawn A$32.56 million, undrawn A$5.41 million.

Item 8.5 of the Appendix 4C requires an entity with negative operating cash flow to state its estimated quarters of funding available. Amaero's answer is 2.28.

The arithmetic is item 8.4 divided by item 8.1: total available funding of A$24.84 million, being cash of A$19.43 million plus undrawn facilities of A$5.41 million, divided by quarterly operating cash use of A$10.90 million.

Two points about that ratio, one in the company's favour and one against.

In its favour: the denominator includes A$2.8 million of non recurring IPO and redomiciliation costs and A$0.4 million of costs related to the May manufacturing incidents. Adjusted general and administrative expense was A$6.0 million in Q4 against A$5.9 million in Q3, so the underlying cost base is flat rather than escalating. Strip the one off items and the burn rate looks materially better.

Against: the 2.28 figure counts operating cash use only. It excludes investing activities entirely. Amaero spent A$9.38 million on investing in the same quarter. Total cash consumption, operating plus investing, was A$20.28 million against A$24.84 million of available funding. On that basis the company has closer to one and a quarter quarters, not two and a quarter, unless capital expenditure stops.

Capital expenditure will slow, because the three year program is complete. But the fourth atomizer and the argon recycler are both ordered.

There is one more wrinkle. The investor presentation shows a different set of cash numbers than the 4C, and the difference is not an error in either document.

The presentation shows A$38.3 million at 31 March and A$25.2 million at 30 June. The 4C shows A$33.4 million and A$19.4 million. The gap is restricted cash: A$4.9 million at March and A$5.8 million at June, held under the terms of the Export-Import Bank facility. The footnote states that restricted cash "becomes available to fund principal and interest payments once Amaero achieves a 1.2x debt service coverage ratio."

So the restricted balance is not available to fund operations. It is collateral. The A$19.43 million in the 4C is the spendable number and it is the one I have used throughout. Anyone quoting A$25.2 million as Amaero's cash position is quoting a figure that includes A$5.8 million the company cannot spend.

The presentation also shows a further A$4.7 million EXIM disbursement for previously incurred capital expenses expected in July 2026, which takes pro forma cash to A$29.9 million on the inclusive basis, or roughly A$24.1 million on the spendable basis.

On the debt itself: the EXIM Bank credit agreement was signed in February 2025 at US$22.8 million and amended in June 2026 to US$26.1 million, an increase of US$3.3 million providing US$2.9 million of additional net cash after the exposure fee. A fixed rate of 5.36 percent was locked at execution, giving a total effective rate of approximately 8 percent. Interest payments commence 30 September 2026 and are quarterly thereafter. Principal repays in twenty eight quarterly instalments beginning 30 September 2027 and ending 30 June 2034. The facility reflects a 75 percent loan to cost advance on capital equipment at the Tennessee facility.

Note the date. Interest payments begin next month. Principal begins in thirteen months. This is non dilutive capital, which is genuinely valuable, but it is capital that starts demanding cash back on a fixed schedule starting almost immediately.

Put the whole picture together and the S-1 stops looking like a strategic option and starts looking like a requirement. A company with roughly two quarters of operating runway, capital equipment on order, interest payments starting next month, and a stated intention to list in late calendar 2026 or early 2027 is a company that needs the listing to happen roughly on schedule.

That is not a criticism. It is the single most important thing to understand about the timeline, and it cuts both ways. It means management is highly motivated to deliver contract news into the registration window. It also means the company has limited ability to wait out a bad market.

Who owns this

The FY2025 annual report discloses 690,670,634 fully paid ordinary shares at 30 June 2025, up from 542,912,442 a year earlier, with issued capital of A$122.6 million against A$77.4 million. Roughly 148 million shares were issued during FY2025 raising approximately A$45 million. FY2026 financing activities brought in a further A$74.6 million.

The current figure is established. Amaero Inc. was admitted to the ASX official list on 11 June 2026 with 953,327,200 CDIs quoted, equivalent to 23,833,180 fully paid ordinary shares at the 40 to 1 ratio. The pre scheme count in Amaero Ltd was 952,933,460 shares.

At the ASX quote of A$0.245, 953.3 million CDIs implies a market capitalization of approximately A$234 million, or roughly US$161 million at prevailing rates. That is the number to work with, and it is derived from a filing rather than a data vendor.

It also frames the dilution history plainly. Amaero listed in December 2019 with 77,073,938 quoted shares and 174,853,651 in total. It arrives at its US listing attempt with 952,933,460. The share count rose monotonically for six and a half years. There was never a consolidation, which means every one of those raises landed on existing holders undiluted by any offsetting mechanic.

The 40 to 1 CDI ratio is the exception, and the scheme booklet is candid about what it is. It describes the exchange ratio as effecting an "implicit consolidation." Functionally it is a forty to one reverse split executed through a redomiciliation, which is how a security priced at twenty four Australian cents becomes a security priced near seven US dollars without a separate consolidation vote.

The ownership structure is concentrated and you should know the shape of it.

Pegasus, the fund led by chairman and chief executive Hank Holland, has led three capital raises for Amaero and owns approximately 28 percent of fully paid ordinary shares, rising to approximately 42 percent on a fully converted, fully diluted basis. Holland's disclosed personal interests at the FY2025 report were 196,291,635 ordinary shares and 171,451,503 options over ordinary shares.

That option position is large enough to require thought. A holder with 171 million options against a base of 690 million shares represents meaningful potential dilution, and the gap between 28 percent outright and 42 percent fully diluted quantifies it. The favourable reading is that the chief executive's fund has repeatedly underwritten the company's capital needs and is aligned with a much higher share price. The unfavourable reading is that a single holder controls a blocking position today and would control something close to a majority on full conversion.

Both readings are correct. Which one dominates depends entirely on what happens to the share price between here and the option expiries, and I do not have the full exercise price and expiry schedule.

One small data point on incentive alignment worth recording: on 2 August 2026 the company filed an Appendix 3H notifying that 166,590 options had ceased because the conditions attaching to them "have not been, or have become incapable of being, satisfied." Some performance hurdle was missed. It is a tiny number of options and I would not build anything on it, but it is evidence the performance conditions are real rather than decorative.

The fires

On 13 May 2026, a flash fire occurred during manufacturing operations at the Tennessee facility, resulting in injuries to two production team members.

I am not going to soften that. Two people were hurt.

The company states the incident was contained to a small area with no damage to the facility or manufacturing equipment. Subsequently, an unrelated flash fire occurred in the facility's exhaust system, with no injuries and no damage.

Amaero's response was to engage Jensen Hughes, a safety and risk based engineering consultancy, to conduct a comprehensive review of facility design, engineering controls and process safety, and to pause titanium powder production for approximately six weeks. Refractory powder production and PM-HIP manufacturing continued throughout, including completion of a A$4.6 million refractory powder order for Titomic. Titanium production resumed on 9 July 2026.

Two facts from the recovery are worth stating because they are the ones that tell you whether the damage was commercial as well as physical. During the six week pause there were no purchase order cancellations and no employee attrition. The financial impact was A$1.3 million of revenue deferred into Q1 FY2027 and A$0.4 million of direct incident related cost.

Fine metal powder is combustible. Titanium powder is more than combustible, it is pyrophoric in the right particle size distribution and oxygen concentration, and flash fire is the characteristic failure mode of powder handling operations. This is a known hazard class with established engineering controls, and the fact that a powder producer had a powder fire is not evidence of unusual negligence.

What I would want to know, and cannot determine from public filings, is whether OSHA opened an inspection, what it found, and whether the Jensen Hughes recommendations have been fully implemented or merely accepted. An injury accident at a facility that is simultaneously seeking to expand its role in the naval nuclear supply chain is not a small governance matter, because the customer in question audits its suppliers.

I will file a records request and report what comes back in a later part.

Testing the claim

Amaero describes itself, in the boilerplate at the foot of every announcement, as "a leading U.S. domestic producer of high-value refractory and titanium alloy powders."

I want to test that sentence, because the difference between "a leading producer" and "the only producer" is the difference between a good business and a monopoly, and small cap narratives have a habit of drifting from the first to the second.

It does not survive as a monopoly claim. It survives comfortably as written.

6K Additive, based in North Andover, Massachusetts, produces C103 spherical powder using its UniMelt microwave plasma process. It was selected by America Makes for the Powder Alloy Development for Additive Manufacturing project, part of a US$6 million programme funded by the Air Force Research Laboratory, specifically covering C103. That is a directly competitive US producer with directly competitive government backing.

Elmet Technologies produces C103 niobium alloy powder for laser powder bed fusion and is a fully integrated US company with facilities in Maine, Ohio and Michigan. AP&C, part of the Colibrium Additive business, lists niobium C-103 in its powder catalogue.

So Amaero is one of at least three or four producers of C103 powder with a United States footprint. The company's own wording is careful and accurate: "a leading" producer, not "the only" one. If you encounter research on this name that claims sole source status for C103, that research is wrong, and you should discount everything else in it accordingly.

Where Amaero does look genuinely differentiated is not in the alloy but in the combination. It runs refractory and titanium atomization at production scale, and it runs PM-HIP, in one facility, in the United States, with a naval nuclear prime already partway through qualifying it. I have not found another company with that specific stack. The moat, if there is one, is the qualification position and the second business line, not the chemistry. The chemistry is in the literature and has been since the Johnson administration.

There is one more competitive note worth recording. Trade press reports that Amaero completed qualification of a C103 powder for ADDMAN, which suggests the customer list extends beyond the names in the ASX announcements. I have not verified this against a primary source.

What has actually been signed

Because the distinction between announced, contracted, and delivered is where small cap theses go to die, here is the ledger in one place.

Read the type column rather than the value column. Exactly one line in that table, the A$7.8 million master purchasing agreement, carries a stated minimum commitment. The Velo3D headline is a demand estimate. The Department of War contract is a real signed development programme with milestone payments. The BPMI award is a real production order for a very small amount of money attached to a very large stated requirement. The UPM agreement is a distribution deal with a 4,000 kilogram opening order and a contracted minimum inventory.

That is a more solid book than most companies of this size have. It is not the same thing as A$35 million of contracted Velo3D revenue, which is how this company is most often described.

What would break this

I have tried to write the bear case as if I were short.

The financing risk is first and it is not subtle. Two and a quarter quarters of operating runway on the company's own disclosure, less than that on any measure that includes capital expenditure, interest payments beginning 30 September 2026, principal beginning 30 September 2027, and a capital raise that is contingent on an SEC review process the company does not control. If the IPO window closes, the alternative is another dilutive placement into an ASX market that has already absorbed A$74.6 million of financing in one year. The FY2025 pattern of roughly 148 million new shares in a year gives you the shape of what that looks like.

Customer concentration is second, and it is worse than the headline suggests. The largest publicised relationship is with a company that has never reported positive gross margin, carries going concern language in its own risk factors, did not file the agreement as a material contract, and does not mention Amaero anywhere in its annual report. The most strategically important relationship, BPMI, has so far produced US$344,000 of production revenue. The gap between the strategic story and the invoiced revenue is wide in both directions.

Third, revenue lumpiness. A business that went 4.7, 3.1, 2.6, 7.8 across four quarters is a business where a single delayed shipment moves the quarter. The company is guiding to a growing backlog, but A$12.9 million of that backlog is scheduled into a single half year and any slippage lands visibly.

Fourth, and this is the one I think is most underrated, the difference between capacity and utilization. Amaero has built 680 tonnes per year of installed capacity and sold A$18.1 million of product. Fixed cost absorption at that utilization is poor, and the company does not disclose gross margin in the quarterly, which is itself worth noticing. Until we see a gross margin line, the operating leverage argument is an assertion.

Fifth, hafnium. The Department of War is paying Amaero to reduce reliance on foreign raw material sources. That is the government telling you the input chain has a problem. Amaero has not disclosed secured hafnium or niobium supply arrangements, and a powder producer without secured inputs is a converter exposed on both sides.

Sixth, the safety record. Two injured workers in May, two fires, an independent engineering review, and a customer base that audits suppliers.

Seventh, dilution and control. Forty two percent fully diluted for a single holder aligned with the chief executive.

Eighth, and least quantifiable, the possibility that hypersonics demand for C103 is a narrative rather than a purchase order stream. C103 is an outstanding vacuum nozzle material with sixty years of flight heritage. Its atmospheric oxidation behaviour is a genuine limitation for sustained airbreathing hypersonic flight, and the coating is a life limiter. A programme funded specifically to develop alternatives to it is not an unambiguous endorsement of its future volume.

What would make it work

The bull case does not require heroic assumptions, which is what makes it interesting.

It requires the BPMI relationship to progress from low rate initial production to production rates against a stated requirement of up to 400 components per year. It requires the IPO to complete inside the guided window at a valuation that reflects US defense materials comparables rather than ASX small cap industrials. It requires utilization on 680, eventually 920, tonnes of installed capacity to rise against a capital base that is already spent. And it requires the FOCI work to convert into the classified contracts the company has now twice told us it is pursuing.

Notice that three of those four are things the company has already substantially de risked. The capital is spent. The redomiciliation is done. The audit is done. The submission is in. What remains is conversion.

The asymmetry, stated plainly: this is a company whose strategic position is materially further along than its income statement, trading on an exchange where its natural buyer base cannot reach it, with a catalyst that mechanically fixes the second problem inside twelve months.

The risk, stated equally plainly: it has to get there on about two quarters of cash.

Where I stand

I do not have a position in Amaero and I am not initiating one today.

The reason is not the thesis. The thesis is the most interesting one I have looked at this quarter. The reason is two unresolved facts and one execution problem.

The first unresolved fact is gross margin. Amaero does not disclose it quarterly. The entire operating leverage argument, which is the argument for owning a company with 680 tonnes of capacity and A$18.1 million of revenue, rests on a line item that has not been published. I can see revenue and I can see cash burn. I cannot see what the revenue earns before overhead, and at low utilization that number can be almost anything.

The second is raw material supply. Amaero has disclosed no hafnium or niobium sourcing arrangements. In a market where the input has risen eightfold since 2020 and world production is roughly seventy five tonnes, a converter without disclosed supply security is carrying a risk I cannot size.

The execution problem is the relationship between the two tickers, and it is the practical thing most readers will care about.

The ASX quote is A$0.245. The American over the counter line, AMROF, last printed at US$0.235. At prevailing exchange rates those do not reconcile. A$0.245 converts to roughly US$0.17. AMROF is quoting something like a thirty five percent premium to its own primary listing.

That gap is far too wide to be settlement friction, and the explanation is almost certainly that AMROF is thin enough that its last print is stale, wide, or both. This is the standard trap in dual listed micro caps. The ASX line is where the price is set, by Australian investors, during Australian hours, in Australian dollars. AMROF is a derivative of that price and its quote is only as good as its last trade.

If you are going to own this, own it knowing that. Watch 3DA, convert at spot, and treat any AMROF print above that as an offer rather than a price. Never pay through the primary line. The one structural comfort is that this arbitrage has a defined end date, because a US listing collapses the two quotes into one.

I would rather be early on a company like this than right about a company like this, so let me say plainly what would move me from watching to owning: a published gross margin above roughly thirty percent, disclosure of secured hafnium supply, or the S-1 going public. Any one of the three.

What I want before I act: the Form S-1 when it goes public, which resolves the share count, the margin structure, the customer concentration disclosure and the risk factors in one document. It also, being an SEC registration rather than an ASX release, will be written by lawyers with a materially different liability exposure.

The catalyst calendar between now and then, as best I can construct it: FY2026 annual report and audited accounts, expected September 2026 based on the prior year's 18 September filing. First interest payment on the EXIM facility, 30 September 2026. Q1 FY2027 quarterly and Appendix 4C, expected late October 2026, which is the first clean look at post incident production and will contain the A$1.3 million of deferred titanium revenue. The public flip of the S-1, expected on the company's own guidance in late calendar 2026 or early 2027. Completion of the Department of War programme, August 2027.

The one I would watch hardest is the October quarterly, because it will contain the first cash figure after the capital programme closed, and it will tell you whether the burn rate normalizes now that the one off IPO and redomiciliation costs are behind it. Item 8.5 in that document is the single most informative number this company publishes.

Part 2 will cover the S-1 when it becomes public, the hafnium and niobium input chain in detail, the full BPMI contract history if I can reconstruct it from federal procurement records, and the OSHA record on the May incidents.

I have no position in Amaero, Velo3D, or any company named in this piece. This is not investment advice. The company is loss making, has negative operating cash flow, and states on its own filing that it has approximately 2.28 quarters of funding available.

References

Amaero Inc., Quarterly Activities Report and Appendix 4C for the period ending 30 June 2026, ASX announcement 22 July 2026.

Amaero Inc., Investor Presentation, ASX announcement 21 July 2026.

Amaero Inc., Board Update, ASX announcement 4 August 2026.

Amaero Inc., Final Director's Interest Notice, Jamie Levy and Alistair Cray, ASX announcements 4 August 2026.

Amaero Inc., Notification of cessation of securities, Appendix 3H, ASX announcement 3 August 2026.

Amaero Inc., Submission of Draft Registration for IPO, ASX announcement 15 July 2026.

Amaero Inc., Amaero Awarded A$6.5 Million Contract from U.S. Department of War, ASX announcement 22 July 2026.

Amaero Inc., Piping Contract Awarded from BPMI for Submarine Industrial Base, ASX announcement 1 July 2026.

Amaero Inc., Amaero Resumes Titanium Powder Production, ASX announcement 9 July 2026.

Amaero Inc., Amaero Comments on Incident at Factory, ASX announcement 15 May 2026.

Amaero Inc., Update on Tennessee Manufacturing Facility, ASX announcement 27 May 2026.

Amaero Inc., Amaero Expands Powder Capacity Ahead of Schedule, ASX announcement 22 June 2026.

Amaero Inc., Amaero Re-domiciliation, Schemes become Effective, ASX announcement 11 June 2026.

Amaero Inc., United States Navy Issues Letter of Support, ASX announcement 15 December 2025.

Amaero Inc., Amaero Signs Supplier Agreement with A$7.8M Order for Titanium Powders, ASX announcement 10 April 2026.

Amaero Inc., Credit Agreement Signed for US$22.8 Million Loan from Export-Import Bank, ASX announcement 26 February 2025.

Amaero Ltd, Annual Report 2025, ASX announcement 18 September 2025.

Velo3D Inc., Form 8-K and supplementary slides, SEC accession 0000950170-25-070439, filed 13 May 2025.

Velo3D Inc., Form 8-K and supplementary slides, SEC accession 0000950170-25-104039, filed 6 August 2025.

Velo3D Inc., Form 10-K for fiscal year 2025, filed 31 March 2026, commitments and contingencies note.

Velo3D Inc., Form 10-Q for the first quarter of 2026, commitments and contingencies note.

Velo3D Inc., SEC company submissions record, CIK 0001825079.

Amaero, Amaero Signs Long-Term, Exclusive Supply Agreement with Velo3D, announcement 29 April 2025.

Velo3D Inc., Velo3D Announces $22 Million Long-Term Exclusive Supply Agreement, announcement 29 April 2025.

Amaero, Velo3D Qualification Condition Satisfied for C103, announcement 8 July 2025.

Amaero International Ltd, prospectus dated 29 October 2019, sections covering Monash intellectual property licence agreements.

Amaero Inc., Scheme Booklet, distributed May 2026, including CDI ratio and implicit consolidation disclosure.

ASX admission notice, Amaero Inc., 11 June 2026, 953,327,200 CDIs quoted.

United States Geological Survey and industry trade estimates for hafnium production, concentration and pricing, 2024 to 2026.

6K Additive, niobium alloy powder C103 product documentation and America Makes PADAM award announcement.

Elmet Technologies, C103 niobium additive manufacturing powder product documentation.

Amaero investor archive, investorhub.amaero.com.au, 659 announcements reviewed.

Not a chip. A crystal. Someone grows a boule of indium phosphide, slices it into wafers, polishes them flat enough that a laser structure can be grown on top, and ships them to the epitaxy houses. Those epiwafers become lasers. Those lasers become 800G and 1.6T optical modules. Those modules are the only reason a GPU in one rack can talk to a GPU in another rack fast enough to matter.

The market has spent two years pricing the modules. It has spent about two days pricing the crystal.

On Thursday July 30, after the close, AXT reported a quarter that broke a three year pattern. Revenue of $47.6 million against roughly $34.8 million of consensus, up about 165 percent year on year. Gross profit of $21.4 million. Operating profit of $10.4 million. The stock rose 27 percent that session and traded up another 24 percent after hours.

I want to do three things. Walk through what the quarter said. Then explain the one mechanism that actually determines AXT's reported revenue, which is not demand and which almost nobody modelling this company has right. Then put the price into the context that a two day chart hides.

What the quarter said

Revenue has moved in a step, not a drift. Six quarters of history and the guided September quarter, which the company put at roughly $66 million on the call:

The margin line matters more than the revenue line, and it is the reason this is worth your time at all.

Fifteen months ago this business sold product below cost. Gross margin was negative 6.4 percent. It printed 44.9 percent in the June quarter. That is a fifty point swing in five quarters and none of it came from a price increase. It came from volume moving through a fixed cost base.

A crystal growing operation is close to a fixed cost business. The furnaces run whether they are full or not. The difference between running them at sixty percent and ninety five percent is most of the income statement. Which is why the bottom line flipped so violently:

One correction to the headline before we go further. Diluted EPS was $0.17, but core EPS is closer to $0.11. About a third of pre-tax income is treasury yield on the cash from the April raise, not operations. Interest on a bank balance is real money and it is not the business, and it will not scale with wafer volume.

A second one. Minority interests take 14.6 percent of everything from here. AXT does not own all of Tongmei, the Chinese subsidiary that grows the crystal. When you read consolidated profit, roughly a seventh of it belongs to somebody else.

And a third thing you cannot find in any filing. AXT's filed segment disclosure has exactly two buckets, Substrates and Raw materials and other. The product level split exists only on the earnings call, and it shows how concentrated this company has become:

Two thirds of AXT is now one product line. If you are reading the 10-Q you cannot see that.

The mechanism nobody is modelling

Here is the part that changes how you read every AXT quarter, including this one.

AXT's reported revenue is gated by Chinese export permits, not by customer demand.

On February 4, 2025, China's Ministry of Commerce issued Announcement No.10, placing indium phosphide substrates under dual-use export licensing as control code 3C004.a, alongside trimethylindium, triethylindium and the associated production technology. Every shipment of indium phosphide out of China to a customer outside China now requires a permit. Permits run roughly sixty business days per order.

Watch what that does to the revenue line:

That is not a demand curve. Demand for AI optical interconnect did not fall 39 percent in the fourth quarter of 2025 and then rise 70 percent in the first quarter of 2026. Permits did.

The company said so directly. Fourth quarter revenue came in at $23.0 million against guidance of $27 to $30 million, and the chief executive explained the miss in these words: "we didn't receive as many export permits in Q4 as we had hoped based on the average processing time we had seen up to that point in October."

So when a $47.6 million quarter lands against $34.8 million of consensus, the honest reading is that it is at least as much a permit release event as a demand event. Permits granted inside a quarter convert to revenue inside that quarter. The backlog they unlock is lumpy by construction.

Two consequences follow and they point in opposite directions.

The bearish one. A single administrative decision in Beijing can produce another fourth quarter style air pocket with zero change in end demand. If you own this stock you are underwriting a permit office.

The bullish one, which is subtler and which I think is underappreciated. If revenue is permit-gated rather than demand-gated, then reported revenue is a floor on demand, not a measure of it. The orders exist whether or not the paperwork cleared. And the air pockets are mechanical, which means they are buyable.

One detail that softens the risk. In the first quarter of 2026, 61.5 percent of revenue went to China-destination customers, which requires no export permit at all. Only 38.5 percent was licence-exposed. Management guides a 40 to 60 percent China mix. The permit exposure is real and it is not the whole company, and the mix is a lever management can pull.

Three things about the trade regime that are not in the coverage

One. There is a 50 percent tariff on this product and it was filed under solar.

HTS 3818.00.00 is the classification for doped indium phosphide and doped gallium arsenide substrates. Effective January 1, 2025, the US Trade Representative moved that entire eight digit subheading from the 25 percent Section 301 list to the 50 percent list. The action was framed publicly as a solar measure covering polysilicon and wafers, but it was executed at the eight digit level rather than the statistical suffix level, so compound semiconductor substrates were swept in as collateral scope.

AXT's product is Chinese origin. Any Chinese origin doped indium phosphide or gallium arsenide substrate entering the United States has carried a 50 percent duty since the start of 2025.

That also explains a number which looks alarming without context. North America revenue was $1.708 million in fiscal 2025, down 77.5 percent. AXT states plainly that it cannot forecast when export permits will allow indium phosphide shipments to the United States, and that no gallium arsenide permit to the US has ever been approved. The American market is functionally closed to them right now.

Two. The November 2025 truce did not touch any of this. Announcement No.10 was deliberately left in force. Only Article 2 of the separate Announcement No.46 was suspended, and that suspension runs to November 27, 2026. The indium phosphide regime is legally distinct from the gallium and germanium regime and it has no expiry date.

Three. The 25 percent Section 232 semiconductor tariff does not apply. Effective January 15, 2026, its scope is logic integrated circuits in HTS 8471.50, 8471.80 and 8473.30. A search of the full Federal Register notice returns zero hits for 3818, wafer, substrate, indium or gallium arsenide. Substrates escape it entirely. That is a bear talking point that turns out to be wrong, and knowing which risks are real is most of the work.

The demand is corroborated, and this is the strongest part of the file

If revenue is permit-gated, you need independent evidence that the underlying orders are real. There is some, and it is unusually hard evidence. Customers are prepaying for capacity.

The Lumentum agreement runs to December 31, 2031 and carries a second $43.5 million tranche in 2028.

That is $91.2 million of customer cash committed in a year when the company will do somewhere near $190 million of revenue. Firms do not wire deposits of that size for capacity they are speculating about. Two of the three counterparties are among the largest optical component makers in the world and have every ability to source elsewhere if they thought they could.

This is the single most persuasive item in the bull case and it is more persuasive than the revenue print, because a prepayment is a decision made by a customer with its own money at risk.

Against it, management's stated 2027 exit target is roughly $130 million per quarter of indium phosphide capacity, about $520 million annualised. That number exists only on an earnings call and requires roughly $100 million of 2027 capital spending to actually happen. Treat it as an aspiration with a funding condition attached, not a forecast.

The price, which is the part the two day move hides

Here is where I have to correct something I wrote earlier this week.

The obvious way to describe this stock is that it rose 57 percent in two sessions and is therefore extended. That is what the tape looks like if you start on Wednesday. It is also wrong.

At $58 this stock is 59.4 percent below its May high and still 9.5 percent below the price at which the company itself placed equity in April. The 57 percent move is a bounce off a deep drawdown, not a breakout to new highs. Anyone who bought the April placement is underwater.

That is a materially different setup from the one the two day chart suggests, and it changes the answer to whether you are chasing. You are not buying a stock making new highs. You are buying one that halved and is trying to come back on a genuine change in the numbers.

What it does not change: insiders sold roughly 266,000 shares between $86.73 and $115.24 about six weeks before this print. They were not selling into the blowout. They were selling before it, at much higher prices. That is not a crime and it is not proof of anything. It is a fact worth having.

What $58 is asking for

Enterprise value at $58.16, adjusted properly for the cash and long-term investments that screeners miss, puts AXT around 25.5 times trailing sales. That is above Lumentum at 21.7 and far above Coherent at 7.5.

On the September quarter run rate of $66 million annualised, the same enterprise value is about 12.1 times sales. At the $46.94 close it is 9.3 times.

So the valuation question reduces to one thing. Does the guided quarter happen, and does it hold. At 25 times trailing you are paying for a business that no longer exists. At 12 times forward you are paying a reasonable multiple for something growing triple digits, if the growth is real and if it recurs.

Given that revenue is permit-gated, whether it recurs is not a question about demand. It is a question about paperwork.

What I am doing

I am starting a position of 15,000 shares. That is a starter, not a full weight.

The plan is to accumulate on weakness, and the permit mechanism is precisely why that plan fits this stock. If reported revenue is gated by an administrative process running on a sixty business day cycle, this company will produce quarters that miss badly for reasons that have nothing to do with its customers. Those are the entries. A business whose air pockets are mechanical rather than fundamental is a business you want to buy on the air pockets.

What I will not do is put full size on before the September quarter tells us whether $66 million was a permit release or a new baseline.

What Part 2 has to answer

Does the guided $66 million land, and what was the permit count behind it. Everything else is downstream of that.

The capacity build. The April raise was $632.5 million against a company doing $27 million a quarter at the time. I want the furnace count, the cost per unit of capacity, the qualification timeline, and when incremental capacity actually produces revenue.

Six inch indium phosphide, which is the cost curve. Where is yield running, what is the customer qualification timeline, and is AXT genuinely ahead of Sumitomo and JX or merely first to announce.

The Tongmei listing on the Shanghai STAR Market. A separate Chinese listing of the subsidiary that owns the crystal growing is the largest structural question a US shareholder faces here. How much would be sold, and what is left for us.

The 2028 supply wave. Sumitomo is targeting 3.1 times its 2024 indium phosphide capacity. San'an has a six and a half billion yuan six inch line. SMART Photonics is bringing an Eindhoven fab to full operation. If all of that lands, the pricing window that produced 45 percent gross margins closes. I want the supply and demand curves side by side with dates on them.

Customer concentration. Three counterparties just prepaid $91.2 million. That is excellent evidence of demand and poor evidence of diversification.

Part 2 carries the model. This one is the map.

New position disclosed above, opened around the time of publication. Not investment advice. Every figure is sourced below and dated.

References

• AXT, Inc. Announces Second Quarter 2026 Financial Results, 30 July 2026 (Business Wire; SEC 8-K EX-99.1, accession 0001437749-26-025061)

• AXT, Inc. Announces First Quarter 2026 Financial Results, 30 April 2026

• AXT, Inc. Form 10-K fiscal 2025, SEC CIK 0001051627, accession 0001437749-26-008612, Note 14 geographical information

• AXT, Inc. Form 10-Q first quarter 2026, accession 0001437749-26-017054

• US Trade Representative, Section 301 Federal Register notice modifying subheadings, effective 1 January 2025, Annex B, HTS 3818.00.00

• Federal Register, Section 232 semiconductor proclamation implementation, 20 January 2026, US note 39

• MOFCOM and General Administration of Customs Announcement No.10 of 2025, 4 February 2025, control codes 3C004.a through 3C004.c and 3E004

• Semiconductor Today, AXT's Q4/2025 revenue constrained by delay in China export permits, 9 March 2026

• Semiconductor Today, AXT's revenue grows 17% in Q1 after greater-than-expected export permits, 5 May 2026

• Capacity reservation agreements, AXT 8-K filings and press releases, Lumentum, Coherent and Nanjing Casela

• Form 4 filings, accessions 0001437749-26-019464 and 0001437749-26-019347

• Price and volume data, 29 and 30 July 2026, NASDAQ consolidated tape

The clean version goes like this. China restricted germanium exports. Germanium is the material every infrared lens is made from. LightPath Technologies makes BlackDiamond, a germanium-free glass, and is therefore the only way for Western defense programs to keep building thermal imagers. Buy the stock.

Two of those sentences are true. The one in the middle is not, and LightPath says so itself, in its own annual report, in writing.

What is left after you remove it is a more interesting company than the story implies, sitting on top of two disclosures made in the last nine days that nobody has written about.

First, the part of the story that is wrong

BlackDiamond BD6 is arsenic selenide, As40Se60. That is not a proprietary formula. It is a well known chalcogenide composition.

VITRON Spezialwerkstoffe in Jena, Germany publishes a glass called IG 6. Its composition is As40Se60. The same material. SCHOTT sells the same family as IRG 22 through IRG 27. Umicore sells GASIR.

You do not have to take my word for it, because LightPath's own fiscal 2025 annual report says it plainly:

"Some of these materials such as BD6 and BD2, are generic, and are produced by at least two other companies. The crown jewels of our materials, the BDNL type, are exclusive to us."

That is the company telling you the material at the center of the popular thesis is generic. Any bull case resting on BD6 being the only germanium substitute is going to meet that sentence eventually, and it will lose.

So what is actually defensible? Four things, and they are better than the chemistry claim because they are harder to copy.

Domestic melt capacity. Making the glass in the United States is not the same as knowing the formula. VITRON is German, SCHOTT is German, Umicore is Belgian. For a US defense program with domestic sourcing requirements, where the glass is melted is the qualifying fact, not what is in it.

A government-funded qualification path. This is the piece I find most persuasive and it is completely absent from coverage. The Defense Logistics Agency has paid LightPath twice, under two sequential contracts, specifically to qualify chalcogenide as a defense material:

Material qualification is the actual barrier to entry in defense optics. Not the recipe. The recipe is in the literature. What takes years is getting a material onto a qualified list and into a program of record, and the US government is underwriting LightPath's path through that process. That is public, dated, and durable in a way a glass composition is not.

Molding yield. Chalcogenide is precision moulded rather than diamond turned. That process knowledge is real and it does not transfer with a datasheet.

Vertical integration into cameras, via the G5 Infrared acquisition, which is where most of the value and most of the accounting complexity now sits.

And the honest technical caveat. Chalcogenide is not a straight drop-in for germanium. Transmission band and thermal drift behaviour differ. Some designs substitute cleanly and some require a redesign. Anyone telling you it is a simple swap has not read a datasheet.

The disclosure from nine days ago that nobody has read

On July 22, LightPath filed twenty two post-effective amendments and seven S-8 post-effective amendments. On July 23, twenty two effectiveness notices came back. Twenty two of anything in one day gets my attention.

Most of it is housekeeping. The cover pages say removal from registration of securities that were not sold.

But buried in it is something live. Registration number 333-262768, filed in February 2022, registered $75.8 million of shelf capacity plus $25.2 million under an at-the-market sales agreement with A.G.P. and Alliance Global Partners. Deregistering it formally terminates that at-the-market facility.

That matters more than it sounds. An at-the-market programme is a machine that sells stock into the market quietly, every day, at whatever the bid is. Turning it off means future capital raises have to be visible, discrete block deals rather than a continuous drip. For a company with LightPath's history of returning to the equity market, replacing silent supply with announced supply is a structural improvement in how shareholders get treated, and nobody has noticed it happened.

The remaining live vehicles are the December 2025 primary shelf and two resale shelves.

The disclosure from eight days ago that is genuinely uncomfortable

On July 23, LightPath agreed to sell one hundred percent of its Chinese subsidiary, LightPath Zhenjiang Optical Instrumentation, to a company called Hengtu Optical Technology in Nanjing.

Read the counterparty description from the filing carefully:

owned by the Purchaser Representative (Mr. Leo Zheng) and certain members of the Company's current management team.

LightPath's own management is on the buying side.

The price is $4.5 million paid over five years, minimum $500,000 per year, with 4 percent financing interest per installment and 7 percent on default, unsecured on the public record. LightPath retains a board observer seat and licenses over the trademarks for five years along with product drawings, tooling, moulds, process documentation, technology and know-how. There is a five year exclusive supply agreement, so LightPath keeps buying from the factory it sold.

Those China operations generated roughly $4.5 million per year of third party revenue. So LightPath is selling a business at approximately one times the revenue it gives up, collecting the money at $500,000 a year, from its own managers, with no disclosed fairness opinion.

Now the clause that made me stop:

"for every payment by the Purchaser of $1,000,000 of the principal... the cost plus markup percentage for products supplied to LightPath shall automatically increase by five percent (5%)... Upon full payment... fixed at cost plus thirty percent (30%)"

The supply agreement starts at cost plus ten percent. Every million dollars the buyer repays raises the price LightPath pays for its own products by five points. Stepped out, it looks like this:

The faster management pays LightPath for the factory, the worse LightPath's gross margin gets.

I want to be fair about this. There is a legitimate strategic reading. A company selling a China-substitution thesis while manufacturing in China has an obvious contradiction to resolve, and selling to the local management team who actually run the plant is a normal way to divest an operation that will not attract an arm's length buyer. Retaining supply, tooling and an observer seat is prudent.

But the ratchet is not standard, the counterparty includes insiders, no fairness opinion is disclosed, and the transaction has not closed yet. Those four facts belong in front of anyone underwriting this equity.

G5 is the company now, and the earnout tells you what it earned

LightPath bought G5 Infrared in February 2025. G5 makes cooled and uncooled thermal cameras. Camera level integration carries better margins than selling lens elements to somebody else who builds the camera.

G5's financial performance is never separately disclosed. But the earnout structure is disclosed, and the earnout can be solved backwards. This is the most useful piece of arithmetic in the file.

Start with the base year. The purchase agreement contained a clawback equal to 50 percent of any shortfall below $17.3 million of calendar 2024 revenue. The annual report shows a revenue clawback of $1,104,471. So $17,300,000 minus twice $1,104,471 gives $15,091,058. G5 did $15.09 million of revenue in calendar 2024. That number appears nowhere in any filing. It falls out of the clawback.

Then year one. The earnout ladder paid $3.5 million above $21 million of revenue, $7.0 million above $23 million, $10.5 million above $25 million and $14.0 million above $27 million. In the quarter ended March 31, 2026, LightPath paid $7.3 million in cash plus 297,445 shares worth $3.2 million. Exactly $10.5 million. That is the third tier.

Then year two, and this is the tell. The ladder pays $4.5 million above $30 million and $9.0 million above $33 million. On April 16, 2026, LightPath signed a side letter fixing year two at the maximum $9.0 million, ten and a half months before the measurement period even ends and four and a half months earlier than the contract required payment.

You do not agree to pay the top tier early unless you already know you are hitting it.

That is de facto forward guidance of G5 revenue above $33 million, disclosed in a contingent consideration footnote instead of a press release.

What it actually cost. All in, G5 is roughly $43.1 million rather than the $27.1 million headline, of which about $32.3 million is cash. The sellers collect $19.5 million of a possible $23.0 million of earnout, or 84.8 percent. That is a business bought well if the growth holds and expensively if it does not.

Why the reported loss is mostly not a loss

The nine month fiscal 2026 net loss is $16,404,698. That is the number a screener shows you and it looks awful for a company this size. Inside it is $12,234,529 of change in fair value of acquisition liabilities, which is the earnout being marked up because G5 keeps beating its tiers.

Strip it out and the underlying loss is about $4.17 million, with an underlying operating loss near $3.12 million. Quarter by quarter, excluding the mark, that loss narrows every single period:

That is a completely different picture from a $16.4 million hole.

Here is the elegant, perverse part. The better G5 performs, the larger the reported GAAP loss becomes, because the earnout liability marks up. Anyone screening this company on GAAP net income is being shown a number that gets worse as the business gets better.

The earnout liability trajectory is the cleanest public scorecard on a subsidiary that is otherwise invisible:

The jump in the December quarter is the moment G5 blew through its tiers.

The thing happening in about 48 hours

The June 3, 2026 offering carried twin sixty day restrictions that expire on or about August 2, 2026.

The first is a lock-up on all directors and executive officers and on North Run Strategic Opportunities Fund I, LP. North Run's residual 6,717,376 shares, 9.6 percent of the company, become saleable.

The second is LightPath's own restriction on issuing stock or convertibles, including the ban on variable rate transactions. The company becomes free to raise again, directly into the fiscal 2026 annual report window, which historically lands in late September.

I am not predicting either happens. I am telling you the doors open this week, and that a 9.6 percent holder becoming free to sell is the kind of thing that later gets attributed to news.

The order book, which is bigger than the press releases and is in the filings

Correction, added 31 July 2026. An earlier version of this piece said the backlog could not be verified from financial statements and that anyone quoting an order figure was quoting a press release. That was wrong, and a paid subscriber who has held the stock over a year and visited the Orlando facility corrected me within hours of publication. LightPath discloses backlog in MD&A as a named key indicator, with a five quarter table. I looked in the wrong part of the document. The corrected section follows, and the real number is much better than the one I used.

From the quarterly report for the period ended 31 March 2026, verbatim:

"Our total backlog at March 31, 2026 was approximately $110.6 million, an increase of 196%, as compared to $37.4 million as of June 30, 2025."

That is $110.6 million of firm orders against a business that did roughly $45 million of revenue in the trailing year. It is not a press release number, it is management discussion in a filed report, and the company defines it tightly: firm orders reasonably believed to remain in the backlog and convert to revenue, including customer purchase orders and qualifying supply contracts.

The composition matters more than the total.

Read the first line again. Fifty eight million dollars of it comes from "a leading global technology customer" for advanced infrared camera systems shipping in calendar 2026 and 2027. Not a defense prime. A technology company. I had assumed the large camera orders were a prime contractor buying for a military programme, and the filing says otherwise. Whoever that customer is, they are more than half the backlog, which is both the strongest fact in the bull case and the single largest concentration risk in the company.

There is also a third acquisition I had not found. LightPath closed on Amorphous Materials LLC on 21 January 2026. Amorphous Materials makes chalcogenide glass. So the company whose own annual report concedes that BD6 is generic went and bought one of the other producers of it. That is a direct answer to the objection I raised earlier in this piece, and it is a better answer than anything in the marketing.

And in the engineering services line, the filing names a Visimid development contract with Lockheed Martin. That is a real prime relationship, disclosed, currently small enough that its timing moves the segment around.

On the federal contract data, the original point still stands and is worth keeping. LightPath has approximately $1.0 million of lifetime federal prime contracts across six awards. It is a subcontractor. Its product goes into somebody else's camera. Prime to subcontractor purchase orders are not public, so USASpending will never show you this book. The mistake I made was concluding that therefore nothing was verifiable. The company reports it directly; you just have to read the management discussion rather than the contract database.

One research note for anyone following this trail. A plain search for LIGHTPATH in federal contract data returns about thirty awards, but most of them belong to a completely different company, the Lightpath fibre business selling internet circuits to the DEA, the FBI and the Maritime Administration. Filter to LIGHTPATH TECHNOLOGIES or you will overstate the defense footprint by roughly sixty percent.

Where I come out

I am not taking a position today. Here is the honest reason.

The company is more interesting than its story. The real moat, domestic melt plus a government-funded qualification path plus moulding yield plus camera integration, is more durable than the germanium narrative that got people here, and buying Amorphous Materials in January strengthens it further. G5 is compounding fast enough that management pre-committed to its maximum earnout. Backlog is $110.6 million against roughly $45 million of trailing revenue and it is disclosed, not asserted. The underlying loss is a quarter of the reported one and shrinking. The at-the-market facility is off.

Against that, a related party sale of the China factory to management with a margin ratchet and no fairness opinion, a 9.6 percent holder coming unlocked this week, a fresh ability to issue equity in the same window, more than half the backlog resting on one unnamed customer, and a central marketing claim the company's own annual report contradicts.

Those are not disqualifying. They are the reason to do the second half of the work before putting money down rather than after.

What Part 2 has to answer

Does the China sale close on these terms. No completion has been filed. Terms can change and the market has priced none of this.

Fiscal 2026 full year numbers. The year ended June 30. The annual report lands around late September and will carry the first clean full year view with G5 and AML in it, plus the June backlog figure.

Whether North Run sells. Ownership filings after August 2 answer that quickly.

Segment margins, camera versus component, separately. The entire vertical integration thesis rests on cameras being structurally better and nobody has shown the split.

Who the customer is. One unnamed leading global technology customer accounts for roughly $58 million of a $110.6 million backlog. Identifying it is the single highest value open question in this name, and the fact that it is a technology company rather than a defense prime narrows the field considerably.

Where the glass gets melted after the divestiture, and what the cost plus ratchet does to gross margin at each repayment step, modelled out.

What Amorphous Materials actually brings. LightPath bought a chalcogenide glass producer in January. Price, capacity, customer list, and whether it closes the generic-material gap or simply adds volume.

And the competitive question that decides the thesis. If VITRON and SCHOTT already make the same composition, what stops a US licensee of either from qualifying domestically. The answer is probably the qualification timeline. I want that timeline quantified in years.

Part 2 carries the model. This one is the map, and the map has a few roads on it that are not on anyone else's.

No position at time of writing. Not investment advice. Every figure sourced below and dated.

References

• LightPath Technologies Form 10-K fiscal 2025, SEC CIK 0000889971, accession 0001654954-25-011130, Item 1 and Note 3

• Form 10-Q, quarter ended 31 March 2026, accession 0001437749-26-015594, Note 4 and segment reconciliation

• Form 8-K, 23 July 2026, accession 0001437749-26-024234, Item 1.01 and Exhibit 99.1, China divestiture

• Form 8-K, 3 June 2026, accession 0001437749-26-019435, Item 1.01, offering and lock-up

• Membership Interest Purchase Agreement, G5 Infrared, Exhibit 10.1 to 8-K accession 0001654954-25-001429, sections 2.8(b) and 2.8(c)

• Post-effective amendments, accessions 0001437749-26-024147 through 0001437749-26-024176, and EFFECT 9999999995-26-002408

• Registration statement 333-262768, A.G.P. and Alliance Global Partners at-the-market facility

• USASpending prime award records, recipient LIGHTPATH TECHNOLOGIES, contracts SP800023P0004 and SP800024C0009

• LightPath BD6 datasheet, Infrared Chalcogenide Glass As40Se60

• VITRON Spezialwerkstoffe GmbH IG 6 specification and SCHOTT IRG series datasheets

• Price and volume data, 30 July 2026, NASDAQ consolidated tape

The answer is on the record. It came on July 23 with the half-year report.

The order did not come.

What came instead was a set of numbers almost nobody has read properly, including the financial press that covered them. LPKF has four million euros of cash. It also claims eighty percent of the market it is trying to win. Both of those sentences come from documents published on the same day.

This post is the longest in the series because the research is. I read the half-year report, the earnings call deck, the annual general meeting voting record, the shareholder counter-motions, the chief financial officer's speech to that meeting, the articles of incorporation, and the full German short-selling register going back to 2013. Several of the things I found have not been reported anywhere. Two of the things that were reported are wrong, and I will show you why.

Let's go.

Read more