The Discount Went Extreme While The Demand Started Becoming Law
The stock fell to the cash-floor bear case while the C-130J design review cleared and the FAA moved to mandate the very traffic layer Merlin's autonomy already fuses.
Long $MRLN. It is my largest position. Everything here is built from public data and Shawarma Capital's own model. None of it is leaked or paid for.
Nothing here is investment advice. For educational purposes only. I may hold positions in names I discuss. Do your own research. No liability assumed.
Executive Summary
Since Part 11, two things moved in opposite directions, and the gap between them is the entire setup.
The price went down. On June 3, when Part 11 published, Merlin Labs (NASDAQ: MRLN) closed near $7.60. As I write at the end of June it trades near $5, a touch above its all-time low and roughly at the cash-floor bear case in my own model. Re-anchor to the live price when you read this.
The company went up. In the same window it completed the critical design review on the C-130J autonomy program, the single largest tripwire in this whole series, and it held. Congress kept legislating the demand. And on June 30 the Wall Street Journal reported the FAA is preparing a mandate that would require nearly all aircraft in civilian airspace to use ADS-B, the post-crash safety push turning into rulemaking.
So the business cleared its biggest gate and the addressable airspace started being upgraded into the exact data Merlin's autonomy consumes, while the stock printed its worst month. That is not a contradiction to resolve. It is the trade.
The architecture of this series has not changed. The trade was never the story, the contracts, or the technology. It is one variable, the execution-risk discount, and that variable compresses on milestones the company is already contracted to hit. Part 11 showed two conservative discounted-cash-flow models, mine and TD Cowen's, landing within pennies of an $11 floor, with Cowen's own undiscounted model implying roughly $27. The stock is now near $5. The discount did not narrow. It went extreme.
One thing I am updating in the open, because a paid reader deserves it and because the September and October mechanics are exactly what Part 11 said this Part would mark. The fully diluted share count is higher than the 135 million I used in Part 11, because the 12 percent convertible preferred is large and is likely to convert near a floor price this autumn. On the fully diluted, post-conversion count, my conservative floor is no longer $11.52. It is closer to $8 to $9.50, depending on how you treat the conversion. That is the honest number. The stock is near $5. The floor came down, the stock came down more, and the discount to a conservatively diluted floor is wider today than it was in Part 11, not narrower.
This is what this document covers. The CDR and what it removed. The demand becoming law. The patent moat the incumbents are citing. The four autonomy pools and the clean balance sheet. Then the part that matters most right now, the September lockup and the October preferred reset, modeled straight, with no flinching. Then the competitive clock, honestly. Then scenarios and the kill criteria.
Let's go.
I. The Setup Got Better And The Price Got Worse
Start with the price, because it is the objection.
Merlin came public in a de-SPAC that closed March 16, 2026 at a $10 reference. It ran to about $17, missed a quarter that was always going to be a near-zero revenue print, and bled to roughly $5. That is a 50 percent drawdown from the deal price and about 70 percent from the high. I am not going to pretend that is fun. It is the cost of owning a pre-revenue, cert-gated, supply-overhung small cap in a tape that has no patience for any of those three words.
But read what the drawdown is and is not. It is not a fundamental break. Across the entire public record there is no design-review failure, no contract cancellation, no FAA suspension, no management exodus that touched the certification bench, no runway-breaking cash event, and no adverse action on the preferred. Those are the seven things from Part 10 that would actually end the thesis. As of today, none has happened. What happened instead is a calendar problem meeting a supply problem. The revenue ramp is a 2027 event, the lockup is a September event, and the market is discounting both into a thin float right now.
The series mechanic still holds, and at a lower price it is simply more extreme. If the entire distance between the tape and fair value is one input, the execution-risk discount, then a lower tape on an unchanged business is a bigger discount, not a worse company. TD Cowen's published model is the cleanest illustration. They reached an $11 target two ways that agree, a 15 times calendar-2027 enterprise-value-to-sales multiple and a discounted cash flow of $10.94. And the $10.94 is not their fair value. It is their fair value after a 60 percent execution-risk haircut. Strip the haircut and the same model implies about $27, which is almost exactly where Roth's initial $25 target sat. A skeptical defense desk thinks Merlin is worth roughly $27 if it executes and roughly $11 once you discount for the risk that it does not. The stock trades near $5, below even the haircut number.
The discount is not a permanent property of the company. It is a function of uncertainty, and milestones remove uncertainty. A cleared design review removes it. A first flight removes it. A funded task order removes it. Each one gives a careful analyst a reason to take the haircut from 60 toward 50 toward 40, and the per-share number climbs toward the undiscounted case on every step. You are being paid to wait for a discount to melt while the company does what it is already under contract to do. The melt just got cheaper to buy.
II. The CDR Cleared, And The Contract Under It Is Sole-Source
On June 4, Merlin announced it completed the critical design review on the C-130J autonomy program for U.S. Special Operations Command. Preliminary design review was March 5. CDR is the gate where the design is judged mature enough to build and integrate. It is the milestone that, in Part 10, I named the single biggest near-term tripwire, because a failed CDR would have been a genuine kill signal. It cleared. The stock popped about 25 percent on the day and then gave it all back in the broader bleed, which is the discount mechanic working in reverse, the milestone landing and the market refusing to re-rate. That refusal is the opportunity.
Here is the part the deep work this month sharpened, and it is genuinely new to this series. The $105 million USSOCOM vehicle is not an ordinary contract. It is structured as an SBIR Phase III. That matters for two concrete reasons. Phase III is the one SBIR mechanism with no dollar ceiling and no competition requirement, reserved for work that derives from a prior Phase I or II. So it is sole-source justifiable, which means a competitor cannot simply underbid Merlin off this vehicle. And it legally requires a predicate, which Merlin has, an earlier AFWERX Phase II effort. The chain is AFWERX Phase II into a USSOCOM Phase III. That is a clean, defensible procurement posture, not a press release.
Be precise about the money, because the bull case is not served by inflating it. Of the $105 million ceiling, roughly $17.7 million is obligated to date, most of it on the C-130J task order. The money that funds it has come as congressional adds, $10.5 million in fiscal 2024 and a separate $12.5 million in fiscal 2026, the latter visible in SOCOM's own budget justification as funds being added to an existing contract with Merlin. Those adds are non-recurring. They have to be re-won each cycle, which is exactly why the lobbying spend pivoted to fiscal 2027 in the first quarter. So the bull statement is narrow and true. The flagship contract is a sole-source Phase III with about $89 million of unfunded headroom that SOCOM can draw against without a recompete, the program just cleared its design gate, and the customer briefs it as a real line of effort. The bear statement is also true. The dollars in hand are modest and the next tranche is an appropriations event, not a sure thing. Both are in the price.
And the customer confirms the program independent of Merlin's own marketing, which is the validation a skeptic should demand. The USSOCOM Program Executive Office for Fixed Wing publicly briefs MC-130J automation and reduced crew workload as a funded line of effort in its cleared conference materials, and it formalized a Mission Autonomy and Applied AI line this year. The slides do not name Merlin. The budget justification does. Put those two public documents together and you have a real customer requirement with Merlin's name on the funding line. That is more than most pre-revenue defense names can show.
III. The Demand Just Started Becoming Law
Now the ADS-B headline, read correctly, because the lazy version of this is wrong and the correct version is better.
The lazy version says a new ADS-B mandate means thousands of planes need upgrades, so Merlin wins. That is wrong. Merlin does not sell ADS-B hardware. The equipage dollars from any ADS-B rule go to the avionics box makers, Garmin and L3Harris and uAvionix and Honeywell. The market knew this instantly, which is why the original report carried Garmin and L3Harris cashtags and not Merlin's. ADS-B Out has also been mandatory in U.S. controlled airspace since January 1, 2020, so a post-crash rule is really about ADS-B In, receiving traffic, and about closing the transmit exemptions that let the January 2025 Washington midair happen with a helicopter that was effectively not squawking. This is an equipage story for the hardware vendors, full stop.
The correct version is the one that should interest a Merlin owner, and the patent work this month makes it concrete. A nationwide mandate that forces a denser, more complete ADS-B traffic picture is upgrading the airspace into the exact data Merlin's autonomy already eats. Merlin's detect-and-avoid intellectual property does not just consume one sensor. Its directed-perception patent application explicitly fuses ADS-B, air-traffic-control audio, and vision into a single cued detect-and-avoid pipeline, refining the picture with ADS-B and even generating a negative-contact report when it confirms no traffic is present. An autonomous aircraft is only as safe as the traffic picture it can build. A federal rule that makes that picture richer, more universal, and more reliable is a structural tailwind to every system that has to see and avoid without a human, and Merlin holds granted and pending IP precisely on fusing that richer picture into a flight decision.
So state it cleanly and do not oversell it. ADS-B equipage revenue is not a Merlin line item. The denser mandated traffic layer is free fuel for Merlin's autonomy, it strengthens the safety case the FAA will judge Merlin's certification against, and it puts a federal regulator on record that the post-crash answer is more automation and more shared situational awareness, not less. That is the rising tide. It is real, it is on-thesis, and it is not a contract.
The demand that is a contract is the federal budget. The fiscal 2027 request carries large autonomy lines, and Congress has been writing reduced-crew and autonomous-systems language into draft defense bills. The honest caveat is that most of the biggest autonomy dollars, the Collaborative Combat Aircraft money and the missile-defense pools, are aimed at uncrewed fighters and kill chains, not crewed transports, and Merlin has to compete inside the narrow lane that fits its product. But the lane exists, it is funded, and the lobbying machine pushing the fiscal 2027 add is a sophisticated one. Watch whether the C-130J add re-ups. That single line is the cleanest revenue-reality test in the whole thesis.
IV. The Moat Is The IP The Incumbents Are Citing
The strongest single fact this month did not come from a press release. It came from the patent office.
Merlin holds fifteen granted U.S. patents. The core of the moat is the air-traffic-control audio family, seven granted patents plus two fresh continuations filed late in 2024, on parsing the controller's spoken instruction and obeying it, hearing the tower and flying the clearance. The root patent on this is forward-cited by the people who would know whether it matters, and the list is the tell. Honeywell is the single heaviest citer of the ATC-audio IP, and Honeywell is also a stated Merlin partner. Airbus, Boeing, Collins Aerospace, and Thales all cite the family as well. A public avionics peer, Innovative Solutions and Support, cites Merlin's dual-assurance certifiable-compute patent. When the legacy primes who dominate the cockpit are citing your patents in their own filings, you own something real and specific in the one place autonomy has to live, the certified flight deck.
There is a second piece of IP that separates Merlin from its nearest competitor in one sentence, and the CEO put it on the record. Merlin's architecture is true onboard autonomy, with the decision-making software running on the airplane and requiring no link to the ground. That is the structural difference from Reliable Robotics, whose model keeps a remote pilot in the loop. A remote-pilot architecture inherits a datalink as a single point of failure and a regulatory and electronic-warfare vulnerability. An onboard architecture does not. For a defense customer worried about contested, comms-denied environments, and for a certification authority worried about link integrity, onboard autonomy is the harder problem and the more defensible answer. Merlin chose the harder problem, and the patent estate plus the dual-assurance, deterministic-core design is how it intends to make that choice certifiable in a way a black-box system cannot be.
I will mark the honest edges of the moat, because a real one survives them. The patents are commercially untested. There has been no challenge at the patent office and no litigation, so the moat has neither been validated by surviving an attack nor monetized by asserting it. And big technology companies cite the same root ATC-audio patent as generic natural-language prior art, which means the method is broad and the lane around pure airspace autonomy is the defensible part, not language parsing in general. With those caveats stated, the structural read stands. The incumbents who could build this instead chose, in Honeywell's case, to partner, and they are citing Merlin's filings. That is the moat.
V. Four Pools, One Brain, Priced For One
The framing that has anchored this series holds. Merlin is not a cargo company. It is one autonomy core, the Merlin Pilot, that sits inside four separate demand pools, and the market is paying for roughly one of them.
The first pool is military fixed-wing transport, the C-130J under the sole-source Phase III, with explicit scope to extend to other special-operations aircraft. This is the funded, customer-confirmed pool, and it is the one Wall Street's $11 target is mostly built on.
The second pool is the tanker, the KC-135 Center Console Refresh, where Merlin works through GE Aerospace. Be honest about this one. CCR is a real, budgeted Air Force program, with proposals due in mid-September, but its stated scope is a cockpit obsolescence refresh, fuel system and flight displays and the flight computer, not autonomy. Autonomy is the differentiator GE and Merlin are adding on top. And GE has to win it full and open against the entrenched incumbent, Collins, who already protested the prior version into a restart. The GE relationship is a credible door, not a signed distribution deal, and the right way to value it is as optionality on a 2027 award, not as revenue today. The tell to watch is a sub-award showing up under GE's name with Merlin inside it.
The third pool is commercial cargo, branded Condor, the largest pool by far over time and the one Wall Street excludes from its model entirely. This is the free optionality. It is also the longest-dated, because the commercial certification clock is the slow one and the regulator is deliberately cautious. The first revenue here is reduced-crew, a co-pilot's workload handled by software with a human still in the seat, not an empty cockpit. Read the New Zealand 2027 milestone the same way. It is achievable as a reduced-crew, safety-pilot first-of-type, which is genuinely on track, and it is not a pilotless revenue flight, because no jurisdiction has a settled rule for that yet. The optionality is real. The timeline is long. Both are true.
The fourth pool is international, anchored by the exclusive teaming agreement with the Remah International Group to scope autonomy opportunities in the United Arab Emirates. Treat this as a soft, framework-stage memorandum with a genuine, well-connected counterparty, not as booked revenue. It carries no value today and an export-control path that is multi-year. It is a call option on a Gulf market, nothing more, and I am not modeling a dollar of it.
One brain, four pools. The Street pays for the first, hopes for the second, excludes the third, and ignores the fourth. That is the structural mispricing the whole series has argued, and a $5 tape only widens it.
VI. The Balance Sheet Is Clean, And That Is Underrated
The most underrated fact about Merlin right now is that it is boring on the balance sheet, in the way a pre-revenue company almost never gets to be.
It is debt-free. At the merger close, the prior secured loan was repaid in cash from the trust, the older facility was cleared in 2025, and the New Zealand government loan was prepaid in February. There is no maturity wall and no covenant risk. Cash was about $122.8 million at the end of the first quarter, and the company added an $80 million PIPE on May 1, taking pro-forma liquidity to roughly $183 million. Against an operating cash burn near $90 million a year, that is a runway into about mid-2028, comfortably past the 2027 certification and revenue milestones that define the thesis.
The accounting under it is mostly conservative, which I checked line by line and which matters for a name this many investors assume must be playing games. Research and development is fully expensed, with zero capitalized software development, the cleanest possible treatment. Deferred revenue is zero, so there is no channel-stuffing or revenue-pull-forward. There is no related-party revenue. There has been no restatement and no non-reliance filing. The scary headline numbers are mostly optics. The $90 million first-quarter net loss is roughly 70 percent non-cash, driven by the fair-value remeasurement of instruments that have since converted to equity, and the real cash burn was about $23.6 million. The $641 million accumulated deficit is dominated by a one-time, non-cash deemed dividend from the pre-merger preferred exchange, a recap artifact, not cash that was incinerated.
There is one piece of conviction in the capital structure worth naming because it is non-conflicted. Baillie Gifford, an existing holder, did not just hold through the deal. It bought into the PIPE at $12 a share against a trust redemption value near $10.39, a deliberate premium. That is the cleanest single insider conviction signal in the file, a blue-chip long-duration manager paying up rather than redeeming. It does not make the stock go up. It tells you who was buying when it was easy to walk away.
I will state the offsetting accounting flag in the same breath, because a clean section that hides it is not clean. The first quarter printed a positive gross profit only because of a favorable release of a contract-loss reserve. Strip that release and the underlying gross margin is still negative. That is the one number an analyst must normalize out, and I am normalizing it out in front of you.
VII. The September Lockup And The October Preferred, Modeled Straight
This is the section Part 11 promised, and it is the one that matters most between now and the autumn. I am going to model the supply overhang without flinching, because it is the real cost of owning this today and because the discount only makes sense if you can see clearly what it is discounting.
Two mechanics dominate the next four months.
The first is the lockup. Roughly 180 days from the March 16 close lands around September 16. The shares that open are the rollover and sponsor blocks, and the honest count is larger than this series previously carried. Adding the holder I under-counted before, the founder, the sponsor, the venture leads, and the rollover blocks total roughly 56 million shares, against a public float that is a fraction of that. The most motivated seller is the sponsor, holding about 8.8 million shares at a basis near four-tenths of a cent. At any positive price that block is deeply in the money and structurally inclined to sell. The lockup has an early-release trapdoor on a subsequent change of control, but absent that, mid-September is the largest single supply event on the calendar.
The second is the preferred, and this is where I am correcting my own prior model in public. Merlin has about 21.7 million shares of a 12 percent cumulative convertible preferred outstanding, a stated value near $260 million, carried on the balance sheet around $180 million. It accrues at 12 percent paid-in-kind, which is roughly $32 million a year, an amount that is about equal to the entire 2026 revenue plan and several times the actual revenue run-rate. It is senior to the common. And its conversion price, currently $12, is subject to a one-time reset this autumn, on about the twenty-first trading day after the six-month anniversary of the close, so roughly mid-October. If the trailing twenty-day volume-weighted price is below the conversion price, it resets down to the greater of that price and a $5.00 floor. With the stock near $5, that reset is likely to trigger near the floor, and a preferred that converts at $5 instead of $12 converts into far more common, on the order of 50 to 75 million shares rather than 22 million.
Put those together and the fully diluted share count is not the 135 million I used in Part 11. It is closer to 185 to 210 million once the preferred resets and converts and the warrants are counted. That is why my conservative DCF floor is no longer $11.52. Spread the same equity value across the larger count and the floor is closer to $8 to $9.50, depending on whether you net out the preferred claim before or after conversion. I would rather hand you the lower, honest floor than defend the prettier one.
Now the part that keeps this in the bull column rather than out of it. First, the dilution is not new information being sprung on you. It is in the filings, it is the reason the stock trades where it does, and a $5 tape already embeds a market that assumes the reset triggers. You are not going to be surprised by it. The surprise risk runs the other way, if the stock is above $5 into October the reset is less punitive than feared. Second, the conversion is not pure destruction. When the preferred converts it stops being a $260 million senior claim that compounds at 12 percent ahead of you, and it stops draining $32 million a year of seniority. Swapping a senior, compounding claim for common is dilutive to the share count and accretive to the common's position in the stack at the same time. Third, the company is debt-free with a runway into 2028, so it is not a forced seller of equity to survive, which means the supply is mechanical, not desperate. And fourth, every one of these is a dated, known event. Mechanical supply on a known date into a thin float is the textbook setup for a price that is artificially depressed before the event and re-rates after it clears, the same way the lockup expiry that everyone fears so often marks the bottom rather than a new leg down.
So I will say it plainly. The single best reason the stock is at $5 and not $8 is the September and October supply, and the single best reason to be buying the discount is that the supply is known, dated, mechanical, and largely already in the price, in front of a 2027 ramp that is not.
VIII. The Competitive Clock, Honestly
A bull case that pretends there is no competition is a pump. There is competition, and the honest read of it is that Merlin is behind on one race and effectively alone on another.
Reliable Robotics is the credible rival and it is ahead on the nearest visible prize, the first paid uncrewed cargo flight. It raised about $160 million at roughly a $1 billion valuation, it holds the only city-led, uncrewed-cargo slot in the FAA pilot program, and it plans first paid uncrewed freight out of Albuquerque this summer, on the same Cessna Caravan airframe Merlin flies. Merlin is not in that pilot program, and its product is reduced-crew, not uncrewed, so it will not own the "first paid uncrewed" headline this year. That is a real competitive fact and I am not hiding it. The capital gap is real too. Shield AI is valued near $13 billion, Reliable near $1 billion, and the sponsor's next de-SPAC, Elroy Air, is coming at the same $800 million stamp Merlin carried. Better-funded peers are raising at scale while Merlin trades at half its deal price.
Here is why Merlin is not in those races and why that is fine for the thesis. No rival occupies Merlin's exact lane, which is crewed, large, multi-engine aircraft, the C-130J and the KC-135 and the commercial freighters, reduced first and autonomous later, using onboard autonomy and the ATC-audio moat. Reliable is a single-airframe uncrewed Caravan story. The eVTOL field is passenger air taxis. Shield AI is fighters and drones. Those are different, often earlier, often smaller prizes. Merlin's prize is later and larger, the multi-crew heavy-aircraft fleet where the labor savings per tail are the highest and where no one else is running the same play. The KC-135 CCR award is the concrete, funded entry point into that lane, and field-wide certification slippage, even the leader's uncrewed type certificate has slipped toward 2029, means Merlin's 2027 reduced-crew timeline is not uniquely behind. It is on a different, defensible clock.
IX. Scenarios, And The Kill Criteria That Still Stand
The scenario ladder that has anchored this series holds, re-anchored to a tape near $5 and a fully diluted count that now reflects the preferred. The new DCF work raised the floor's honesty, not the ceiling's height.
The honest conservative floor is the diluted DCF, roughly $8 to $9.50 once the preferred resets and converts, which is above the current tape. On top of that floor sits the probability-weighted catalyst optionality the series has carried, which takes the expected value into the mid-teens, and the bull and moonshot rungs are intact because nothing this month lowered the ceiling. Anyone reading this as a low-single-digit story is reading the stack upside down. The price is near the bear rung while the base, bull, and moonshot rungs are unchanged.
And the discipline that keeps this honest is the kill list. From Part 10, seven things would end this thesis. A failed design review. A contract cancellation. A contradiction of the customer-program claims. An FAA enforcement suspension. A management exodus from the certification bench. A runway-breaking cash event. An adverse action on the preferred. As of today, after the deepest audit I have done on this name, not one has occurred. The CDR cleared. The contracts stand. The customer briefs the program. There is no FAA action. The certification bench, the safety-assurance and engineering leadership, is intact even though the founding technical team has rotated out. The runway reaches 2028. And the preferred is converting on its own terms, not being weaponized against the common. The thesis is intact because the things that would break it have not happened, and I will tell you the day one of them does.
The Read
Strip it to the core.
A skeptical Wall Street desk and my own conservative model independently landed on an $11 floor in early June, with the desk's undiscounted math near $27, on a business that has since cleared its biggest design gate. The stock then fell to $5. The fully diluted share count is higher than I previously carried, so the honest floor is now $8 to $9.50, not $11.52, and the stock is still below it. Demand for the core capability is being written into draft federal law, and the FAA is moving to upgrade the entire civil airspace into the traffic picture Merlin's autonomy is patented to fuse. The whole distance between a $5 tape and a model that runs from $8 to $27 to $58 is one variable, the execution-risk discount, and that discount comes down on milestones the company is already contracted to hit.
The only honest costs are timing and supply. The revenue is a 2027 event. The September lockup and the October preferred reset are real, dated, mechanical, and largely already in a $5 price. I would rather buy a known overhang on a known date in front of an unknown ramp than the reverse.
The asymmetry is the spread between what is publicly known to be working, a sole-source Phase III, a cleared CDR, a cited patent moat, a debt-free balance sheet, and what is publicly known to be coming, the autumn mechanics, the fiscal 2027 add, the first flights, and a commercial half the Street values at zero. That spread is the widest it has been in this series, and it widened while the company got better, not worse.
I am long. This remains my largest position. Part 13 will mark the September lockup, the October preferred reset, and the second-quarter print as they resolve.
Shawarma Capital
Disclosure. MRLN is my largest position. 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 and I do not owe you a fiduciary duty. Financial projections, including my discounted-cash-flow and catalyst model, are illustrative model outputs based on publicly available data and my own assumptions. They are labeled conservative and are not guarantees or company guidance. Third-party analyst figures are attributed to their published source and are summarized, not reproduced. Microcaps are illiquid and can lose 50% or more in a single session. Do your own due diligence.