Newsletter · · Ashutosh Agarwal
Defense Spending Nears $3 Trillion as Missile Magazines Empty and Compute Heads to Orbit - Aerospace & Defense Weekly - Week of September 19, 2026
Aerospace & Defense Weekly for the week ending September 19, 2026: dealmakers put global defense spending above $3 trillion and defense M&A near $40 billion, Tomahawk, THAAD and Patriot stockpiles won't be replenished until 2030, Mach Industries laid out its cheap-mass blueprint, Brave Capital explained the neoprime playbook, and SpaceX and Cowboy Space described data centers moving into orbit.
Aerospace & Defense Weekly
Week of September 19, 2026: Defense Spending Nears $3 Trillion as Missile Magazines Empty and Compute Heads to Orbit
The week the numbers finally showed up: $3 trillion in global spending, missile magazines that won't refill until 2030, and data centers packing their bags for orbit.
For weeks this story has run on vision and vibes: founders describing the machines they want to build, investors describing the checks they want to write, generals describing the war they fear. This week, the accountants took the microphone. Two dealmakers sat down and put hard figures on all of it: how much the world is now spending on defense, how fast the money is turning into actual deals, and exactly how many missiles America fired at Iran that it cannot replace before 2030. The picture that emerges is less thrilling than a maiden flight and more useful: it tells you whether the boom is real, and where it is most fragile.
Alongside the numbers, the week gave us the clearest operator statement yet of the idea that keeps reorganizing this whole industry (that in a modern war, the factory is the weapon) and a genuinely strange new frontier: the biggest names in space now openly describe themselves as data-center companies, and are preparing to launch computers, not just cameras, into orbit.
One habit we keep, because it changes how much a claim is worth: we tell you whether a speaker is an operator (someone who builds the machine, runs the company, or lends and invests the money at real risk) or a pundit, analyst, journalist, or former official talking from the outside. Both are useful. They are not the same kind of evidence. A founder describing the drone he just flew, or a lender describing the deal he just funded, is worth more than a commentator guessing at the trend. This week leaned heavily toward the operators, and we'll flag which is which as we go.
1. The Numbers Finally Showed Up: A $3 Trillion World and a Defense-Deal Boom
If you read one thing this week, make it the conversation on Middle Market Growth (September 18), which put actual figures under everything the sector has been feeling. The two guests are dealmakers, not commentators: Chris Abato, a managing director at Merit Investment Bank who spent 22 years at Lockheed Martin doing buy-side deals (so a banker who has literally bought companies for a prime contractor), and James Parker, founding partner of Leonid Capital Partners, a private-credit firm focused only on national security that has done "over $600 million of deals in the space." Treat them as insiders to the money, not builders of hardware, but insiders who see the deal flow before anyone else does.
Abato laid out four trends, and the first is simply the scale of the spending. "Global defense spending is projected to be over $3 trillion by the end of this year. That's a 26% increase over the last three years. NATO allies have committed to spending 5% of their GDP annually by 2035. And U.S. defense spending is up 11% over last year." Europe is a big part of that step-up: "European defense spending rose 14% in 2025 to almost $900 billion." And the U.S. number is set to jump again, to "$1.5 trillion next year, which is 44% increase over 2026," a budget that "includes $300 billion for R&D to support new platforms, new weapons, advanced technologies."
The money is turning into deals fast. "Annual deal values rose to just under $40 billion in 2025," Abato said. "That's up from $24 billion in 2024. 61% increase. The number of transactions also increased by 33% to 465 transactions." And the pace accelerated this year: "In 2026, in the second quarter, we've seen deal values double to almost $13 billion. That's up from $6 billion in the same quarter a year ago." The encouraging part, from a dealmaker's chair, is that the buyers are behaving: "we're not seeing this crazy hubris… the acquirers are looking for accretive deals," meaning deals that add to profits right away, "so what's driving up the valuations is a focus on earnings, not just revenue and not just backlog."
Why are the big prime contractors buying instead of building? Because they can't invent fast enough. "The defense primes and the big companies, they can't innovate fast enough with their internal R&D budgets to respond to the change in threats," Abato said. "So they have to acquire the needed technology that's already proven." He sees the government itself trying to move faster, "bypassing the old stodgy way of doing large procurements with three years of RFPs" (the traditional request-for-proposal process) in favor of "unsolicited proposals" and more small-business research grants.
Parker's contribution was the single most important caution of the week, and it should temper any straight-line enthusiasm. Venture and private-equity funds think in seven- to ten-year horizons; the Pentagon budgets one year at a time. "Capital views the government as a trustworthy partner," he said. "It just doesn't necessarily view it as a reliable one." He also drew a sharp line between two very different markets hiding under the word "defense." On one side, quiet private-equity roll-ups of "specialized manufacturing" and service providers that "don't really deviate from their metrics." On the other, the venture world, where "you've got companies that have little to no government revenue commanding hundreds of millions of dollars of valuation." The open question, in his words, is "what happens when some of these sky-high valuations on the VC side of things come home to roost," when the checks stop and those companies have to justify themselves through a real acquisition or IPO.
Both men reached for the same historical rhyme for what's coming in space, where money is pouring into companies that "may not even have a product yet, yet they've got four or five hundred million dollars worth of investment." Abato's analogy: "This is very similar to the early 20th century when we had the automotive players. There was something like 400 automotive manufacturers in 1910. 20 years later, there were 10% of that, with three having 90% of the market share… there's going to be winners and there's going to be losers." Space, he added dryly, "is a really hard environment."
Why it matters: this is the week the boom got audited. The top-line is genuinely enormous and broadening beyond the U.S.: a $3 trillion world, Europe at nearly $900 billion, deal count up a third to 465, deal value up more than 60%. But the two people watching the money most closely both flagged the same fault line: the government is a one-year-at-a-time buyer funding a business that needs a decade of certainty, and a chunk of the private-market valuations are running well ahead of any revenue. The bull case and the bubble warning came from the same two mouths, in the same breath. Hold both.
2. The Empty Magazines, and the $1,000 Drone That Costs $4 Million to Shoot Down
Inside that same conversation, Abato put hard numbers on the story we've been circling for weeks: the war is emptying the shelves faster than industry can refill them, and the economics of how the West fights are upside down.
On the stockpiles: "U.S. fired a thousand Tomahawk missiles. It'll take current production levels till 2030 to replenish that. Same is true for the THAAD and Patriot interceptors" (the ground-based systems that shoot down incoming missiles), "we've burned over 50% of our stockpile." That is the supply side of what recent weeks framed from the battlefield: America has been spending its most expensive, hardest-to-build munitions in real combat, and at today's factory rates the restocking runs for years.
The demand side is a cost problem that borders on absurd, and it is the single clearest reason the whole "cheap mass" movement exists. "There can be an inexpensive drone that costs like $1,000," Abato said, "and it takes a $4 million interceptor to take it out. So those cost economics have to change." When your enemy attacks with something that costs less than a used car and your only answer costs as much as a house, you lose the math even when you win the engagement.
The bottleneck to fixing this is not desire, it's physical capacity, and Abato was blunt that it cannot be rushed. "Supply chain capacity is a huge challenge right now… it takes time to build a facility, bring in a workforce that can actually respond and build things." His memorable line: "You can't line up nine women and get them pregnant for a month and get a baby." Add an aging workforce ("30% of our aerospace and defense workforce is over 55") and the sole-source problem he worried about for years at Lockheed, where "in some cases you only had one supplier, and if one of the competitors bought that supplier, now you are out." Even commercial aviation is straining against the same wall: Airbus's order backlog "reached over 9,000 aircraft as of May of this year," but "production capacity and constraints can't change overnight."
Why it matters: this is the two-sided munitions story told with figures instead of adjectives. The bullish reading is a multi-year replenishment supercycle: someone has to rebuild every Tomahawk, THAAD and Patriot that got fired, and the factory rates say that runway stretches to 2030. The bearish reading is the $1,000-versus-$4-million math: the West is still mostly buying expensive answers to cheap threats, and the capacity to build either the cheap answers or the expensive restocks is constrained by concrete, steel and skilled hands that take years to add. The demand is unarguable. Whether the industrial base can meet it on time is still the question the whole sector is priced on.
3. The Factory Is the Weapon: A Builder's Blueprint for Cheap Mass
For the clearest operator statement of that "cheap mass" idea, The Aerospace Executive Podcast (September 17) sat down with Nathan Diller, president of Mach Industries and a former career Air Force officer and pilot who previously ran AFWERX (the Air Force's innovation arm) on "a billion dollars a year" trying to bring new companies into defense. He is now on the other side of the table, actually building the hardware, which makes him a genuine operator with a rare inside-and-outside view.
Mach's whole pitch is that modern manufacturing can produce weapons the way Silicon Valley produces electronics: cheaply, quickly, and by the millions. "The Venom aircraft, 71 days from sketch to first flight, a fully 3D-printed fighter-jet-looking aircraft," Diller said, describing what he calls "almost a 200th-century shift that is comparable to the Industrial Revolution." The intent is "to manufacture asymmetric aerospace and defense systems… we say by the millions because we really mean this is a step change in how we think about manufacturing."
He walked through the product line, and it reads like a catalog of the new way of war:
- Viper: a turbojet-powered, vertical-takeoff, one-way attack drone "built initially for Army looking at long-range fires," now through "a series of live fires and international tests."
- Glide: a low-cost precision-strike munition, tested and "flown hundreds," that Mach says it can "produce by tens of thousands per month."
- Stratus: a high-altitude balloon acting as a "pseudo-satellite," carrying communications and sensors cheaply, "at a very high volume."
- DART: a low-cost interceptor built to "defeat the types of drone capabilities that we're seeing still in Ukraine today" and during the fighting with Iran, Mach's own answer to the $1,000-drone-versus-$4-million-interceptor problem.
- ATLAS: a larger, runway-independent maritime strike aircraft, selected by the Defense Innovation Unit together with the Navy's NAVAIR.
The philosophy underneath is a "high-low mix." Diller is explicit that the expensive exquisite platforms don't disappear ("there are literally trillions of dollars invested in capabilities that exist today") but that the cheap, high-volume systems have been the missing half, and the money to build them has only just arrived: "It hasn't been until just recently where dollars have actually started to flow there… it's just been impressive to see how quickly there's real funding starting to flow toward what I think is a more balanced high-low mix."
His deepest point was about where things get built, and it echoes the industry-wide shift back toward owning your supply chain. For decades, he said, the model, "largely driven by Wall Street," was for "America to be the idea factory. We go and find an international partner, in many instances China, to go do the manufacturing." The hidden cost was theft: "the amount of lost IP that happens from prototype to scale manufacturing… was just not understood by America at all." Mach's response is to be "a hardware-first company," designing only what it can "source domestically or produce myself domestically, at scale, at cost," and to do it with people who have done high-volume manufacturing before, hiring "some of the world's greatest talent that has spent time at the Teslas, at the SpaceXs, at the Apples." He tied his optimism to recent policy: he was at SpaceX's Starbase when the Defense Secretary "announced a lot of the new acquisition reforms," and called the combined "movement of the Congress, the movement of the Pentagon, and the movement of industry… literally a renaissance in American manufacturing."
Why it matters: this is the "cheap mass" thesis stated by someone actually pouring the metal, and it makes the abstract concrete. The phrase "the factory is the weapon" means that in a long war the decisive edge is not a single brilliant machine but the ability to churn out good-enough machines faster than the enemy, and Diller's product list (drones by the millions, munitions by the tens of thousands a month, cheap interceptors for cheap threats) is exactly what the numbers in sections 1 and 2 say the Pentagon now needs and is finally funding. The catch is the same one the dealmakers flagged: this only works if the domestic supply chain and workforce can actually be built, and if the funding that "just started to flow" keeps flowing.
4. How the Smart Money Picks Winners: The Neoprime Playbook
If Mach is one builder, Connected With Latham (September 17) is a guided tour of the investors backing dozens of them. The guest is a partner at Brave Capital, which holds one of the deeper defense-technology portfolios around (Anduril, Shield AI, Ursa Major, Chariot Defense, Together AI and more), so an investor with real money at stake rather than an outside commentator. (The host, Latham & Watkins partner Haim Zaltzman, disclosed that he is personally a small investor in one of the companies discussed, Shield AI.)
The organizing idea Brave uses is "decisive technology": backing a company at the exact moment a capability crosses from impossible to real. The cleanest example is Shield AI. For "the last 20-plus years, nearly every drone depended on either GPS to know where it is, or a radio link back to a human pilot," fine in Iraq and Afghanistan, "where GPS was readily available" and "interference wasn't an issue." But in today's fighting, "GPS jamming, communications jamming, that's the first thing adversaries do." Shield AI's product is "essentially a software pilot that lives on the aircraft itself. It lets the aircraft navigate, make decisions, complete its mission, even when GPS and comms are unavailable." The bet was timing: "AI finally got good enough to be the pilot" at the same moment the enemy learned to switch off everything else.
The second example, Chariot Defense, is a neat illustration of borrowing an advance from one industry and pointing it at another. The battlefield is still powered by "noisy diesel generators," and "even to get the fuel needed… it's oftentimes the most expensive part about modern warfare. You're risking lives just to move diesel." Chariot didn't invent a new battery. It "simply redirected" the battery packs that the electric-car and electric-aircraft industries "already spent years, billions of dollars, learning how to manufacture" (lightweight, cheap, energy-dense) into military power generation, at the moment those packs "crossed the energy-density and cost threshold" to replace generators at scale.
The bigger framework the Brave partner offered is what they call "neoprimes," a new generation of prime contractors built around frontier technology. Their defining habit, and the tell for investors, is that "rather than waiting for the government to fund R&D," neoprimes "build what they think operators need beforehand" and get "prototypes into the hands of service members," improving them "through real-world feedback." That is the opposite of the traditional model, in which industry waits for a government requirement before it builds anything.
One last thread connected this conversation to the space section below: the biggest shift in AI, the investor argued, is "that AI isn't just a software story right now. It's become a physical infrastructure story." A few years ago the debate was about "who has the best models." Now "the constraints are power, cooling, data-center capacity, chip manufacturing. The bottleneck has become infrastructure," which is why Brave has also invested in a small modular nuclear-reactor company to feed the AI economy's appetite for "reliable, always-on power." The honest caution, from someone who once thought self-driving cars were a year away: the hard part is rarely the technology. "The real constraints are less technical" and "more around trust, regulation, economics… changing human behavior."
Why it matters: this is the buy-side counterpart to the dealmakers in section 1, how the capital actually chooses what to back. The "decisive technology" lens (invest the moment a capability becomes real) and the "neoprime" playbook (build first, sell to the government second) explain both why the venture valuations Parker warned about are so high and why some of them may be justified: companies like Shield AI and Chariot aren't betting on a future breakthrough, they're commercializing one that just happened. But note the built-in humility: the investor's own warning that adoption always takes longer than the technology, because trust and regulation move slowly. The technology is ready faster than the customer is.
5. Compute Leaves the Planet: The Space Economy Grows Up
The strangest and most forward-looking thread of the week was the same one from two very different rooms: space is no longer mainly about cameras, communications and science. It is becoming a place to put data centers.
The overview came from Cogs of War (September 14), which gathered an operator and two analysts. The operator was Joe Yaffe, chief operating and legal officer of Cowboy Space, a startup "building an American energy grid in orbit." The analysts were Brian Whedon of the Aerospace Corporation, a federally funded research center focused on space policy, and Dan Wald, director of space and AI at Booz Allen Hamilton (which co-produces the show, worth noting, as it makes him a promoter as well as an expert).
First, the size of the prize, and a useful disagreement about how to measure it. Whedon, the policy analyst, put the space economy "somewhere between 400 and 600 billion" dollars, summing "corporate revenues as well as government expenditures," with some projecting "a trillion by the 2030s." Wald, the Booz Allen technologist, argued that number misses the point: the right frame is "the convergence of the tremendous amount of capital flowing into AI and AI infrastructure," which he pegged at "$7 trillion of AI infrastructure build-out, the largest infrastructure build-out in human history." In his telling, space is about to stop being a special corner of the economy and simply become part of it.
Yaffe's company is the concrete version of that idea. Cowboy Space is "building very large orbital data centers… which will contain a megawatt of compute power per launch," and separately "launching satellites to collect solar power in low Earth orbit and beam it to Earth by infrared laser," with a demonstration planned for October using "what we think is the highest-power commercial laser ever operated in space." Tellingly, Yaffe insists "we think of ourselves as a data-center company," not a space company: what the customer buys is "resilient, reliable compute… whether that compute is being done in space or in Northern Virginia or East Texas."
The technical unlock underneath it is a shift from radio to light for moving data. Today's satellite links run on radio-frequency spectrum, a scarce and congested resource; the new approach uses laser (optical) links between satellites and to the ground. The speed difference is staggering: one panelist compared an older radio link "at around two kilobits" to optical links "in the gigabyte scale… thousands of orders of magnitude different." And the sheer scale of what's being planned is hard to absorb. Cowboy Space alone "filed an application for a 20,000-satellite constellation of orbital data centers." For context, Whedon noted there are "roughly 16,000 functional active satellites in orbit" today, up from "about 1,000 that existed back in 2010," and that filed applications now run "in the millions." His forecast: "at least another order of magnitude change in the next 5, 10 years," potentially 100,000 satellites, many of them firing lasers at each other and at Earth.
The biggest operator in the field said the same thing, more casually, on All-In (September 15), which hosted Gwynne Shotwell, president and chief operating officer of SpaceX, as pure an operator as this sector has. Two admissions stood out. First, SpaceX has quietly become an AI business: after absorbing Elon Musk's AI company, "it almost seems… it's as much an AI business as it is a space business, by revenue." A rented-computing business she half-jokingly called "Elon Web Services" is now "tens of billions of dollars per quarter in build-out" with "no drop in demand at all." And Starlink, its satellite-internet service, still has room to run: "our market penetration on Starlink is one and a half to two percent… so there's a lot of headroom there."
Second, Shotwell made the case for putting data centers in orbit, and it is the same logic Cowboy Space is chasing. On Earth, "as soon as someone finds out that a data center is going in, real estate goes from 3,000 an acre to 180,000 an acre," and the wait for permits and electrical equipment runs years: "people are talking about generators in three years. We need compute now." In space, "the real estate is infinite. It's free"; cooling is free because "the radiator looks at deep space, the coldest thing ever"; and solar panels "always face the sun." SpaceX, she said, will "launch AI compute satellites" next year alongside its next-generation Starlink craft.
There was one flash of competitive spice worth flagging, because it captures the pecking order in launch. Asked why SpaceX keeps flying its older Falcon 9 and Dragon capsule, which she likened to a "minivan" next to the "road trip to Mars" of the new Starship, Shotwell noted that customers like NASA still depend on it, and then took a swing at the competition: "Boeing has been paid, I think, probably more than we have been paid, to develop their human capsule… Government paid a lot of money for that. They should be able to use it." The message, barely veiled: the government overpaid a legacy prime for a capability SpaceX delivers more cheaply.
Why it matters: this is where the aerospace story and the AI story become the same story. The AI build-out is running into hard physical walls on Earth (power, cooling, land, permitting, transformers) and the space industry is pitching orbit as the release valve, with free real estate, free cooling and constant sun. Whether that pencils out at scale is unproven and capital-intensive, and the regulatory framework for firing high-power lasers between thousands of satellites barely exists yet. But two credible operators, one a scrappy startup and one the dominant launch company on Earth, are building toward the same 2026–2027 milestones. If they're even partly right, the "space economy" stops being a niche and becomes an annex of the AI economy, and the number to watch is not $626 billion but the multi-trillion-dollar compute build-out looking for somewhere to go.
The Bottom Line
After weeks of vision, this was the week of verification, and the verdict is a real boom with a real crack running through it. The scale is no longer in doubt: $3 trillion of global defense spending, Europe up 14% to nearly $900 billion, a U.S. budget headed for $1.5 trillion, and a deal market that grew 61% to $40 billion and is running twice as hot again this year. The demand is unarguable, too: America can't replace the Tomahawks, THAADs and Patriots it has fired before 2030, and it is still answering $1,000 drones with $4 million interceptors. That gap is precisely the business that Mach Industries wants to fill by making weapons "by the millions," that Brave Capital is funding through "neoprimes" that build before the government asks, and that the primes are scrambling to buy because they can't invent it in time.
The crack is that the buyer and the builder are on different clocks. Capital works in decades; the Pentagon budgets by the year. Some of the private-market valuations are, in one lender's words, waiting to "come home to roost." The factories, the workforce and the supply chains that all of this depends on take years to stand up: "you can't line up nine women and get a baby in a month." And the newest frontier, compute in orbit, is being built ahead of the rules that would govern it.
So the question this newsletter keeps returning to is now answered with numbers, and the answer is: the money is real, the need is real, the technology is arriving, and the binding constraint is no longer imagination or even capital, but whether the industrial base and the government's own machinery can move at the speed the moment demands. This week, the operators building it sounded more confident than ever. The people financing it sounded, quietly, like they were watching the exits. Both are right.