Last Updated on July 7, 2026 by Taya Ziv
Every founder I’ve talked to in the last month suddenly has a defense angle. The cybersecurity guy is “dual-use” now. The drone side project is a “national security platform.” The AI wrapper is “for the warfighter.” I get it. The money is loud, the headlines are louder, and nobody wants to be the one who sat out the biggest venture story of the year.
So let me be the friend at the table who says the quiet part. In defense tech, winning the pilot is not the win. Winning the pilot is often the exact moment these companies start to die.
I know how that sounds. Stay with me, because the money and the graveyard are sitting right next to each other in this sector, and almost nobody sprinting in is looking at the second one.
The number that’s pulling everyone in
Crunchbase just closed the books on the first half of 2026, and the defense number is the kind that makes founders rewrite their decks overnight. More than 14.6 billion dollars went into defense, national security, and law enforcement startups in six months. The old record, the best full year the sector ever had, was 9.6 billion in 2025. So it didn’t just grow. It blew past its own best year ever with half the calendar still on the wall.
And the marquee rounds are enormous. Anduril raised 5 billion dollars in May at a 30.5 billion dollar valuation, and back in March the U.S. Army handed it a ten-year contract worth up to 20 billion. Shield AI pulled in 2 billion. Saronic raised 1.75 billion for unmanned ships. Mach Industries took 300 million for drones. Money that used to flinch at anything with a weapon attached is now running straight at it, because geopolitics got scary and AI made autonomous systems suddenly real.
Then, on July 2, Fortune ran the headline every late-cycle founder should tape to the wall: the defense tech boom has become a bubble, or it’s about to. Crunchbase noticed the same thing from a different seat, reporting that the same VCs who poured the money in are already starting to eye the exits. The smart money is quietly looking for the door while the founders are still hunting for the entrance.
Why the pilot is the trap
Here’s the mechanism, and it’s specific to selling to the government in a way most software founders have never had to think about.
In normal startup life, you sell to a company, they pay you, you have revenue, you grow. In defense, your customer is the U.S. government, and the government does not buy the way a company buys. First you win a prototype competition or a pilot. Everyone’s thrilled. There’s a press release. Your investors forward it around. And then you hit the thing the people who’ve done this call the valley of death: the gap between a shiny pilot and an actual program of record, which is the boring bureaucratic phrase for a funded line item in a real budget.
That gap is not weeks. Procurement for these systems has historically run 18 to 36 months, and plenty of founders get told the real request for proposals is coming “in a couple of years” while their runway is measured in quarters. So you won. You have the logo, you have the press, you have no revenue, and you have investors who signed up for a software growth curve now staring at a government calendar. The success rate for companies trying to cross that valley is, by every honest account, dismally low. One industry leader even put a clock on it: after five years of the Pentagon saying it wants to work with startups, there are maybe two years left before founders give up and the private money walks.
That’s the part the 14.6 billion dollar headline hides. The money is real. The buyer’s patience is not, and neither is yours.
I want to be careful here, because I’m not saying defense tech is fake or that nobody makes it. Anduril made it. But look at how. It didn’t win because its drones were the cleverest things in the room. It won because it raised enough, early enough, to survive years of the procurement grind that quietly starved everyone with a shorter cash runway. Its Dutch counter-drone deal went from signature to operational in under a month this May, which sounds amazing until you realize the reason it made the news is that this basically never happens. The exception got a press release precisely because the rule is a two-year slog. Anduril is what surviving the valley of death looks like, and it costs about 5 billion dollars a round to do it. That is not a template a pre-seed founder can copy.
You can see the same shape everywhere in this market. When most of the money piles into a tiny handful of names, the average check looks incredible and the median founder lives somewhere completely different, which is the exact story I got into when I looked at how 300 billion dollars of venture capital in a single quarter mostly went to about four companies. Defense is that same pattern, just wearing camo.
What this actually means if you’re not Anduril
So what do you do if the defense gold rush is real but the buyer might quietly kill you? A few honest moves.
First, be brutally clear about who signs the check and how long they take. This is the same thing I keep hammering about every business, that distribution beats product and selling is the actual company. It’s just that in defense, the “selling” is a multi-year procurement maze with its own language, and if you can’t fund your way across it, the smartest technology on earth won’t save you. The money in the market is a distraction until you’ve proven you can survive the buying cycle. Which is really just the government-flavored version of the 90-day revenue rule that’s replacing build-it-first MVP culture: find the paying customer and the real path to a signed contract before you fall in love with the product.
Second, watch where the boom is actually landing. A lot of this “defense” money is really autonomous-systems and hard-tech money that found a national-security wrapper, the same migration I traced when robotics quietly ate the venture market and the money left pure software behind. If you’re a software founder duct-taping a defense story onto a normal SaaS product, you’re not really in this trade. You’re cosplaying it, and the people who actually buy in this market can tell.
Third, if you genuinely want in, raise like the valley is real, because it is. Assume no revenue for two years. Assume the pilot is a marketing event, not a sale. Build the war chest, or a commercial revenue line that pays the bills, while the government makes up its mind. The founders who die in defense usually don’t die from bad tech. They die because they budgeted like they were selling to a startup and woke up selling to a bureaucracy.
The line nobody puts on the pitch
There’s a fantasy at the center of every defense pitch right now, and it’s basically Field of Dreams. If you build it, they will come. Build the drone, the sensor, the autonomy stack, and the biggest customer on earth will walk out of the corn with a checkbook.
But the government isn’t a ghost in a cornfield. It’s a budget cycle with a two-year memory and a real tolerance for watching companies starve while it decides. The 14.6 billion dollars is real, and so is the boom, and honestly so is the bubble Fortune is warning about. Just remember that the money went to the companies built to wait. If you’re rushing in because the headline is hot, the pilot you’re chasing isn’t the finish line. For most founders reading this, it’s the start of the part where the clock runs out.
Win the pilot. Then find out whether you can survive winning it.
