In May, a company that did not exist two years ago became a unicorn. Kela, founded in the summer of 2024, raised $200 million at a $1.2 billion valuation, led by Stripes and D1, with Bill Ackman and Eric Schmidt in the round, on top of the $100m. it had already taken from Sequoia, Lux Capital, and In-Q-Tel, the CIA's investment arm. About $20m. in revenue. About 150 employees. An operating system for the battlefield.
That same month, the running tally of tech workers laid off worldwide in 2026 crossed 185,000, and more than half of the layoff announcements named AI as the reason.
Two labor markets, one economy. One is telling its engineers that machines can now do their jobs. The other cannot find enough engineers to build the machines.
The defense-tech story is usually told as a stock ticker. Kela hits a billion. Anduril doubles its valuation to $61b. in under a year. Palantir grows revenue 85% and wires itself into the US Army's next command system. The telling is true, and it misses the structure underneath.
Defense tech is not a company, and it is not even a sector. It is a stack: layer upon layer of distinct technologies, from frontier AI down to bent metal, each layer its own industry with its own economics and its own open doors. The consolidation everyone talks about is happening on exactly one floor. Every other floor is fragmenting, multiplying, and hiring.
The view from the penthouse
The top floor is the one everyone photographs. Palantir posted $1.63b. in first-quarter revenue and secured a foundational role in the data layer of the US Army's next-generation command system, on a team led by its penthouse neighbor. Anduril raised $5b. in May at $61b., and by mid-June secondary trades implied closer to $100b. Kela deployed from a single outpost to a double-digit number of positions along Israel's borders inside a year.
This is the platform layer, and here consolidation is real. Someone is about to become the Android of the battlefield, and the contenders are few. Stop looking anywhere but the penthouse and you reach a reasonable conclusion: The game is over, the giants won, nothing sane is left to fund or join.
The conclusion is wrong. Take the elevator down because every floor is its own gold rush.
One floor below the platforms sits compute. The defense semiconductor market was worth $13.4b. in 2025 and is projected to nearly triple within a decade. In December, the Pentagon stood up a contracting vehicle with a $25b. ceiling to speed American-made military chips into weapons.
Below compute sits navigation and sensing, the floor where quantum quietly stopped being science fiction. SandboxAQ's AQNav navigates by reading the Earth's magnetic field, no satellites required, and has logged more than 450 flight hours across four aircraft types on its way into a Pentagon quantum-sensing program. GPS chaos in the eastern Mediterranean pushed 1,650 ships off their charted positions within a week. Quantum navigation is the layer of the stack that exists because GPS now lies.
Below sensing sits firmware and edge autonomy, where the most consequential military module of 2026 costs less than a pair of running shoes. The TFL-1, a vision module built by the Ukrainian start-up The Fourth Law for under $100, flies a strike drone through its final half kilometer on machine vision alone, no signal, no cable. It is certified by Ukraine's Defense Ministry and fielded with roughly 20 brigades. It does not outperform the million-dollar jammers arrayed against it. It makes them irrelevant for the last 500 meters.
And at street level, bent metal. LUCAS, the American loitering munition that made its combat debut over Iran in January, costs $35,000 and was designed by SpektreWorks, a 14-person company in Arizona now targeting 10,000 units a year. The Pentagon's first drone "gauntlet" this year bought 30,000 drones at roughly $5,000 apiece, from 12 different vendors. Not two primes. Twelve vendors.
There is one more floor, and nobody photographs it: the basement, where things get built. Software iterates in seconds. Production lines move at the speed of poured concrete.
Anduril is pouring a billion dollars into Arsenal-1, a hyper-automated factory in Ohio, because it understands where this market will actually be decided: not by whoever writes the most elegant code but by whoever can turn out 10,000 airframes a year without a single sub-tier supplier of motors or thermal batteries choking the line. The scarcest hire in defense tech right now is not a machine learning researcher. It is a manufacturing engineer.
Software at the top, chips and sensors in the middle, firmware and hardware below, and the factory floor under it all. The penthouse is consolidating. Everything below it is multiplying.
Ask the bubble question out loud
I know the objection, so let us say it plainly: This smells like a bubble. A two-year-old company at $1.2b. on $20m. of revenue, another at $61b. on a fraction of the legacy primes' sales, venture capitalists who discovered patriotism the same quarter their software portfolios stalled. If this were a consumer market, those would be the classic tells.
But bubbles pop when the buyer changes his mind, and this buyer is not a consumer. It is the state, and states are rearming on a schedule set by their fears. At the Ankara summit on July 7, NATO approved more than $40b. over five years for counter-drone defense alone, because the current arithmetic is a losing trade: interceptors that cost millions fired at drones that cost less than a family car.
The Israeli numbers make the same point in miniature, and they are the sharpest version of it. Defense exports reached $19.2b. in 2025, up from $14.8b., a fifth consecutive record. Israeli start-ups working with the Defense Ministry's R&D directorate raised more than $1b. in 2025, against about $150m. the year before. That is not a sector growing. That is a sector being called up.
Are some valuations ahead of themselves? Almost certainly, on the top floor. That is where the froth lives, because that is where the tourists look. And when the shakeout comes, it will not come from demand. It will come from the procurement machine's own clock. A request for 300,000 drones is a question, not a contract.
The Pentagon's flagship fast-track unit still averages about 19 months from solicitation to first delivered article, and the graveyard of defense start-ups is full of companies that died in exactly that gap, between a successful demo and a funded budget line. The casualties will cluster where the pricing does: at the top. That is not where the stack lives.
The open floors are at the bottom
Here is what the penthouse-gazing misses. The barriers to entry at the deep layers of this stack are the lowest in the industry's history. A $100 module sidesteps million-dollar jamming systems. Fourteen people built the drone that flew over Iran.
Since October 2023, more than 130 Israeli start-ups have been folded into the war effort, about half in autonomy and AI, a quarter in sensors. Almost none of them needed a fab, a prime contract, or a nine-figure round to enter. They needed a layer, a use case, and speed.
That is the actual opportunity, and it is a double one. For investors, the top floor is priced for perfection while the floors below are priced like the specialist niches they still are. For the engineer holding a layoff notice that blames AI, the asymmetry is sharper still.
In 2018, some 4,000 Google employees signed a petition against a Pentagon AI project, and the company walked away from the contract. That was the moral consensus of an entire industry: Serious engineers do not do weapons. Seven years later, that industry is cutting people by the hundred thousand, and the fastest-growing employer of technical talent in the West is the one that was taboo.
Human life has become the most expensive asset on the war's stock exchange. The labor market is now running that experiment at scale. Talent, like capital, flows to where it is valued.
The layoff email and the funding round are the same document, read from opposite ends. One economy has decided its people are a cost to be optimized away. The other knows they are the arsenal. The stack is tall, the demand is sovereign, and every floor is hiring.
The question is not whether the door is open. It is how long it stays that way.
Tal Pinkasovich is a defense technology analyst and a major (res.) in the Israeli navy, with more than a decade in the Israeli start-up ecosystem. His work on drone warfare and defense economics has appeared in The Jerusalem Post.