I kept replaying one odd image after the Quantico address: three very different men sitting in a room full of junior officers, asked to sketch the weapons their grandchildren might carry. A launch entrepreneur who just became the first person markets treat as a trillionaire. A drone builder who turned a California workshop story into a prime-in-waiting. A former House Speaker whose credential, depending on who you ask, is a book about technology and enterprise. Out of that room came a name that sounds like a shipping lane and a strategy memo at the same time. Project Meridian. And almost on cue, the aerospace and defense tape kept bleeding.
That combination is what stuck with me. A four-star autonomous command on the drawing board. A budget request that jumps by orders of magnitude. The people who sell autonomy invited to help write the shopping list. Then a sector ETF on track for its longest weekly losing streak since it listed in 2011. If the story were only about demand, the stocks should have cheered. They did not. The gap between the speech and the tape is the part worth sitting with.
What Project Meridian Actually Asks the Pentagon to Do
Secretary of War Pete Hegseth used a sprawling State of the Force talk at Marine Corps Base Quantico to say the quiet part in a formal way. The best forecasts of future conflict, he argued, do not live only inside the building. Project Meridian is meant to look ahead, name the domains that have to be mastered, and point at capabilities the force will need across a generation, not a budget cycle. Elon Musk, Anduril founder Palmer Luckey and Newt Gingrich were in the audience and were named to lead the effort. Oversight sits with Emil Michael, the former Uber executive who now holds the Pentagon’s top technology post.
Coverage of the speech was thin on mechanics. No org chart. No budget line of its own. What did leak into the reporting was a clock: the trio is expected to deliver findings inside 120 days, in a public version and a classified one. That is fast for a bureaucracy that usually measures reform in fiscal years. It is also slow enough for markets to front-run the conclusion, which is exactly what they have been doing.
I’ve found that names like this work best when you translate them into a job description. Meridian is not a weapons program. It is a foresight panel with unusual commercial proximity. Its purpose, in Hegseth’s phrasing, is to identify domains to conquer and capabilities to master. Conquer is a strong verb for a peacetime planning exercise. It tells you the tone of the room more than the legal authority of the group.
The customer is asking two of its most ambitious suppliers what the customer ought to buy next. That can be efficient. It can also be a sales call with a classified annex.
A fair reading of the advisory design
Who sits at the table, and why the seating chart matters
Musk’s SpaceX already handles national security launches and, after a blockbuster June 12 listing, trades as a public company. That IPO is what pushed his paper wealth into trillionaire territory in market shorthand. Luckey’s Anduril has moved from defense upstart to something closer to a prime: a reported $20 billion Pentagon award in March, a $5 billion raise in May at a $61 billion valuation, and software work on Golden Dome alongside Palantir. Gingrich is the odd seat. He is not running a factory. He is there, apparently, because he has written about technology and entrepreneurship. In 2026 that passes for a defense credential in some rooms. I am not sure it should, but the invitation was real.
There is a personal subplot that markets noticed even if junior officers did not. More than a year after the DOGE effort cooled and a very public break with the president, Musk is back in formal advisory clothes. In the same stretch he was at a Trump-Xi state dinner and was expected at a White House artificial-intelligence lunch. At one point the administration had weighed keeping SpaceX out of Golden Dome. Feuds in this town age in strange ways. This one aged into a seat.
None of that is a scandal by itself. Governments have always borrowed outside brains. The tension is structural. Two of the three people asked which capabilities the country must master also run companies built to sell those capabilities. A clean process would wall off product pitches from the forecast. A messy one would let the forecast quietly become the pitch. We will not know which version we got until the 120-day papers land.
The affordable-mass argument, without the brochure
Strip away the personalities and there is a real procurement argument underneath. Legacy industry, in the diagnosis several aerospace analysts have been making for years, builds low numbers of expensive, bespoke assets that are hard to scale when a war burns through them. The counter-model, often called affordable mass, tries to do the opposite: selective vertical integration, common components across product lines, designs simple enough to automate, and factories that can switch products without a five-year retool.
The economics explain the rush. Anduril has talked about a path toward something near a 25 percent company operating margin over time, with roughly 70 to 80 percent of revenue on fixed-price work. Traditional primes live on a different mix, heavier on cost-plus, lower on margin, slower on iteration. Fixed price rewards the builder who guessed right on cost. It punishes the one who did not. That is attractive to a department trying to escape overrun culture. It is also a reason investors have treated the new names as a different species from the old ones.
Analysts who sat with Anduril’s management earlier in the year also pointed to early signs of acquisition reform: multi-year frameworks, open testing, programs such as Drone Dominance. One architect of that model now helps draft the wish list. Perhaps the most interesting aspect is how ordinary that conflict looks once you stop treating it as a morality play. Every industry advises its regulator. Defense is simply the industry where the advice can become an appropriation.
AutoWarCom, Project Agincourt, and a 2027 deadline
Meridian was not the only new label. Hegseth unveiled a four-star Autonomous Warfare Command, already shortened in hallway slang to AutoWarCom, with service-like authorities over drones, artificial intelligence, and command-and-control. Target date for operations: October 1, 2027. Until then, an interim effort called Project Agincourt is supposed to pull the Defense Innovation Unit and the Drone Dominance program into faster loops, pairing operators with entrepreneurs. Existing efforts such as Collaborative Combat Aircraft and the Navy’s MQ-25 unmanned tanker get folded in. He called it the fastest peacetime shift in modern military history. That is a claim, not a measurement. History is full of peacetime shifts that looked fast in the speech and slow in the motor pool.
Still, the direction is hard to miss. Autonomy is being treated less like a gadget and more like a service. Services get budgets, billets, and generals. Gadgets get pilot programs that die in the next continuing resolution.
- A standing command with service-like authority over drones, AI and command-and-control
- An operational target of October 1, 2027, with Project Agincourt as the bridge
- Existing unmanned programs pulled under the same tent rather than left as orphans
- A public bet that adaptation cycles, not platform decades, are the unit of competition
The budget number that should have lifted the whole sector
The money, on paper, was already lined up. The Pentagon’s fiscal 2027 request seeks $54.6 billion for the Defense Autonomous Warfare Group, up from $225 million in fiscal 2026. That is a 243-fold increase. Even by Silicon Valley standards, the multiple is rude. Add a $1.1 billion Drone Dominance push to stockpile 300,000 attack drones priced under $5,000 by the end of 2027, and an Army award of $500 million for first-person-view systems to Neros over the summer, and the spending path looks one-way.
Drone stocks soared when the stockpile story first circulated in May. That rally is a useful memory. Demand narratives in this sector have a half-life. They spike on the announcement, then fade when Congress, peace talks, or a rival product intervenes. May was the spike. Autumn has been the fade.
| Item | Figure cited around the address | What it signals |
| Autonomous warfare group request | $54.6 billion in FY27 vs $225 million in FY26 | A category shift, not a tweak |
| Drone Dominance stockpile | $1.1 billion toward 300,000 sub-$5,000 drones by end-2027 | Volume over exquisite platforms |
| FPV award | $500 million to Neros | Smaller builders can clear the gate |
| Meridian clock | Findings expected within 120 days | A near-term catalyst, public and classified |
| AutoWarCom | Operational target October 1, 2027 | Institution, not a pilot |
Tables like that are why a casual reader expects a bid under every ticker with a rotor on the homepage. The tape disagreed. Which brings us to the part investors actually live in.
Why the sector puked anyway
The SPDR S&P Aerospace and Defense ETF, XAR, was down about 22 percent from its August 14 peak by the middle of the Quantico week, in a bear market on that measure, and on track for a seventh straight weekly decline. That would be the longest losing streak since the fund launched in 2011. Year to date it was off roughly 4.1 percent while the S&P 500 was up almost 12 percent. A sector that spent the winter at a premium to the broad market had slipped to a discount. Bernstein’s Douglas Harned put the swing in round numbers: from about a 15 percent premium in February to about a 12 percent discount.
Three explanations keep showing up, and none of them require a conspiracy.
First, peace risk. On September 22, reports surfaced that Iran had offered to reopen the Strait of Hormuz in exchange for an easing of U.S. military pressure. The president floated a negotiated end at the United Nations. The diplomatic track did not land cleanly, but the downward momentum in defense shares was already in the price. Markets price the possibility of quieter seas faster than they price a new command.
Second, the budget is stuck. Neither chamber had passed appropriations. A stopgap through December 11 would freeze spending at 2026 levels. A 243-fold request is a press release until it is an appropriation. Harned’s line was blunt enough: the budget is stalled. Investors who bought the out-year story are being asked to wait through another continuing resolution. Waiting is not a strategy most momentum accounts enjoy.
Third, the midterms. UBS’s Allyson Gordon has said investors remain reluctant to add exposure until after the elections, and the betting markets on Senate control are part of why. Whoever holds Congress holds the pen on appropriations. A sector this dependent on a single customer cannot ignore the customer’s board of directors.
I keep coming back to a simpler version. The speech described a future order book. The stocks trade the next two quarters of cash and the odds that Congress funds it. Those are different instruments. Confusing them is how people buy the top of a drone rally.
Haves, have-nots, and a market that already picked
The damage was not even. A late-August note from a major bank’s industrials desk called defense tech a tale of haves and have-nots, and warned that day-to-day trade in a space-related vehicle was spilling into drone names. The six weeks after that note proved the split. Pure-play drone names were crushed: Red Cat down about 42 percent, Kratos about 34 percent, Unusual Machines about 28 percent, AeroVironment about 26 percent. SpaceX and Palantir were each up roughly 8 percent from the sector peak. The market had already chosen its winners. The largest of them now has a chief executive at the Meridian table.
That split is the trade, if you are honest about it. The haves are priced as platforms: launch, software, data, a seat near the requirement writer. The have-nots are priced as product cycles: one airframe, one award, one congressional mood. At 40 percent off, the have-nots can look interesting. They can also look like value traps if Meridian’s report narrows the lane to a handful of integrated vendors.
AeroVironment is a useful middle case. Analyst notes this month pointed to funded backlog up 37 percent year over year and 23 percent sequentially, with exposure to faster-growing slices of the defense budget. Demand was cited for the LOCUST directed-energy counter-drone system, including a first international order above $50 million, and for the Freedom Eagle interceptor as a lower-cost alternative to traditional missile defense. A company can have a real backlog and a bad stock at the same time. Backlog is a promise. The multiple is a mood.
Rough scoreboard since the August peak: Broad defense ETF: about -22% Red Cat: about -42% Kratos: about -34% Unusual Machines: about -28% AeroVironment: about -26% SpaceX and Palantir: about +8% each
Numbers like these move. Treat them as a snapshot from the week of the speech, not a covenant. The shape matters more than the second decimal: platform names held up, pure-play drones did not, and the index sat in between, dragged by the old primes and the new casualties at once.
Behind the meter, now in uniform
Buried in the same speech was a power plan that energy investors will recognize. Hegseth wants every major U.S. installation to generate its own electricity, so a base keeps running if a cyberattack takes down the grid. The chosen tool is the nuclear microreactor. First unit targeted for September 2028, with Eielson Air Force Base in Alaska leading the Air Force effort.
This did not appear from nowhere. The Army launched a Janus microreactor program last October. In August it picked five companies under agreements worth up to $2.2 billion: Antares at Fort Bragg, BWXT at Fort Campbell, General Atomics at Fort Hood, Radiant at Fort Benning, and Westinghouse at Fort Drum. The Air Force has its own pairings at three bases and has worked with NANO Nuclear. AFWERX gave that company a $1.25 million award last September to study placing its KRONOS microreactor at Joint Base Anacostia-Bolling, then a follow-on in July to adapt the design for installations more broadly.
Readers who have argued that data centers need their own behind-the-meter generation will hear an echo. The Pentagon got there through the threat of a grid-killing attack rather than a hyperscaler’s power bill. Same hardware logic, different fear. Small modular and micro designs remain the only serious long-duration answer to a power crunch that batteries and diesel cannot cover at base scale. Whether the 2028 date holds is a separate question. Nuclear schedules slip. Military ones slip with a salute.
The rest of the speech, because markets trade the sideshow too
Hegseth did not stop at drones and reactors. General and flag officer billets are to be cut by 20 percent, double last year’s 10 percent trim, with reductions due by January 1, 2027. A new domestic base, the first in four decades, was branded a Next Great American Base for more than 15,000 troops, in what he called timeless classical architecture, with states invited to compete to host it, even as large troop cuts in Europe stay on the table. An Office of Religious Affairs will report directly to the secretary. New corps-of-cadets partnerships were named with Hillsdale, Liberty, LSU, Mississippi State and Tuskegee. He accused U.S. media outlets of treason over coverage of the Iran war, weeks after a House member forced a vote on impeaching him. And he drew a cultural line in language that does not need repeating to be understood: fitness, appearance, and a narrow idea of who counts as a warrior.
Most of that is not an earnings driver. Some of it is a political risk factor. Officer cuts change who buys. A new base is a construction and services story if a state actually wins it. Rhetoric about the press and about who may serve is a headline risk for contractors who sell to allies with different domestic politics. I would not build a model on the cadet partnerships. I would keep them in the file as a signal of which institutions the current leadership trusts.
How a 120-day report can move a multiple
Assume the public Meridian paper is competent and the classified annex is where the real targeting lives. What could it say that changes a stock?
- It could bless affordable mass as the default design rule, which helps builders who already iterate on common parts and hurts builders whose margin lives in bespoke sustainment.
- It could name domains, such as counter-drone, attritable air, autonomous logistics, or space-based sensing, and quietly starve others.
- It could recommend open architectures, which sounds neutral and often favors the software layer over the metal.
- It could push fixed-price and multi-year buys, a gift to firms that can forecast cost and a tax on firms that cannot.
- It could stay vague, in which case the speech was theater and the tape was right to yawn.
The classified version is the one program offices will read. Investors will trade the public adjectives and the leaks. That is an ugly information gap, and it is normal in this sector. If you cannot live with it, there are cleaner industries.
A practical way to think about the split
I do not have a price target, and I am suspicious of anyone who produced one from a speech. What I do have is a sorting hat that has worked better than theme slides.
Ask who gets paid if the requirement is rewritten in 120 days. Platform companies with launch, software, or data already inside the fence have a path even if a specific airframe loses. Single-product drone names need the report, or the next award, to mention their lane. Primes with huge legacy backlogs can survive a narrative winter and still disappoint if the growth budget is redirected. Counter-drone and low-cost interceptors sit in an awkward sweet spot: they benefit if affordable mass wins, and they benefit if the threat keeps multiplying cheap attack drones. That is why names with directed-energy or cheap-interceptor exposure kept showing up in constructive notes even as the group fell.
Then ask what has to go right outside the building. Appropriations. A Senate map that does not freeze the request. A Hormuz story that does not turn into a lasting peace premium. None of those are under Meridian’s control. A brilliant forecast with no bill is a white paper.
Risks that do not fit on a drone slide
Conflict of interest is the obvious one, and it is not theoretical. If the public report reads like a product catalog for two companies in the room, oversight bodies will notice, allies will notice, and the next administration may unwind the seating chart. That unwind risk sits on the haves as much as the have-nots. A seat at the table is an asset until it is a headline.
Execution is the second. Stockpiling 300,000 cheap attack drones is a manufacturing and training problem, not a keynote. Attritable systems only work if units will actually expend them, if software updates do not brick a battalion, and if electronic warfare does not make the cheap option useless on week two. Autonomy commands fail in boring ways: data rights, spectrum, logistics, and a sergeant who does not trust the feed.
Politics is the third. A 20 percent cut in general officers changes culture slowly and news cycles quickly. Accusations of treason against newsrooms, and a blunt filter on who may serve, will be litigated outside the acquisition office. Contractors do not need to share the rhetoric to live with the volatility it creates. International orders, the kind AeroVironment just printed, can cool if partner capitals decide the brand risk is not worth the kit.
Valuation is the fourth, and it is the one retail investors skip. A stock down 40 percent is not cheap if the next twelve months of revenue were priced for a budget that is now frozen. Discount to the S&P is a relative fact. It is not a buy signal. February’s premium was not a sell signal either, until it was.
What I would watch between now and the report
Continuing-resolution language, first. A freeze at 2026 levels delays the 243-fold request no matter how stirring the Quantico lines were. Any anomaly that carves out drone or autonomy money is a tell.
Award flow, second. Drone Dominance task orders, counter-drone internationals, and whether Neros-style FPV deals repeat. One award is an anecdote. A cadence is a market.
Peace headlines, third. Hormuz, Iran, and any European troop-cut confirmation. Defense multiples breathe with the threat board. They do not wait for the official history.
The Meridian leak cycle, fourth. A public summary that names domains will be traded in an afternoon. Silence will be traded too, as evidence the panel was decorative.
Microreactor milestones, fifth, for anyone using the speech as an energy story. Site work at Eielson, Janus agreement conversions, and whether the 2028 date survives contact with licensing. Power for bases is a real industrial theme. It is also a slow one.
A longer view, past the seventh down week
Step back from the ticker and the reorganization is coherent, even if the casting is uncomfortable. The department is trying to stop buying a few exquisite systems and start fielding mass, software, and on-site power. It is cutting flag-officer overhead while adding a command for machines. It is inviting outsiders to say what the next generation of kit should be, and it is asking Congress for a budget that would make the invitation matter. That is a strategy. It is not yet a cash-flow statement.
The market’s refusal to buy the whole basket is not, in my view, a verdict that the strategy is fake. It is a verdict that timing, politics, and competitive position are doing more work than the slogan. When the budget moves, or the threat board flares, or the midterms clarify who writes the checks, the haves are already obvious. The have-nots, marked down hard, become interesting only if the 120-day paper leaves them a lane. If it does not, the discount was a warning, not a gift.
There is a metaphor I cannot shake. A meridian is a line you navigate by, not a destination. Project Meridian will draw a line. Ships still have to be funded, crewed, and pointed. Some of the vessels in this convoy are new and fast. Some are old and heavy. A few are small enough to capsize if the line is drawn an inch to the left. The speech drew attention. The appropriation, and the report, will draw the line.
Demand narratives in defense have a half-life. They spike on the announcement, then fade when Congress, diplomacy, or a rival product intervenes. The work is knowing which fade is a sale and which fade is the new price.
If you own the group, the useful question is no longer whether autonomy is coming. The speech answered that in public. The useful question is whether your specific company is a have, a have-not, or a prime waiting to find out which list it landed on. One hundred and twenty days is not long. For a sector on its seventh down week, it may feel longer than it is.
