Ethan Thornton dropped out of MIT at the age of 19 to pursue weapons manufacturing. First, a hydrogen-powered system he prototyped with parts from Home Depot and Amazon didn’t work — “Hydrogen was a bad bet in general,” he told me at TechCrunch last week. The StrictlVC incident in Los Angeles. Three years later, his company, Mach Industries, is running a six-arm programme and, earlier this month, closed a $300 million Series C round at a valuation of $1.8 billion. The startup has now raised about $485 million in total.

Thornton grew up in Burnet,

Texas, a town of about 6,500 residents, is in a family with deep military ties. Around 2017 or 2018 – when he was still in his teens – he began, by his account, to be “really, really concerned” about the rise of China and what he saw as an impending major-power conflict. That concern eventually turned into a conviction that unmanned systems would redefine warfare and that the US was moving too slowly to meet this moment.

What that looks like, in practice, in mid-2026, is six simultaneous weapons programmes and a company that has a lot to prove rather than focusing on one thing, getting it right, and then expanding. Thornton knows that Mac’s diffuse focus creates some lingering questions for outsiders. “It’s very hard,” he said, volunteering Thursday night. But he doesn’t think defence rewards the kind of single-minded focus that rocket launches require. “It’s a chess game you’re playing with an opponent,” he said, adding that “hundreds of different products have to be shipped if we want protection.” Pick just one, he suggested, and you’ve already lost the game.

These are not ordinary products. MAC is working on a vertical-takeoff strike aircraft; a long-range anti-ship missile; two stratospheric systems; an inexpensive surface-to-air interceptor designed to shoot down drones; and – announced earlier this week – a 40-foot, nearly 4,000-pound Navy logistics-and-strike aircraft that flies almost vertically and can range more than a thousand miles with a payload of a thousand pounds. Takes flight.

That last flight is a significant advance for a company whose largest plane to date has been nearly 13 feet long. And none of the six are yet in full-rate production. Thornton says MAC has won about 13 government contracts, most of which are in the middle stages of defence procurement – ​​past preliminary design and in testing at the government threshold but far below the rate-manufacturing level that fewer than 10 industry-wide programmes have ever reached.

He says that we should see operational deployment of several systems by the end of this year, and his goal is to push three of the six into rate manufacturing in the same window — which would mean going from hundreds of units per month to hundreds of thousands in a factory that Thornton says Mack plans to stand up soon.

It’s an aggressive timeline placed on top of already aggressive stakes. But Mac’s underlying thesis is that the US can’t produce more than China, so it has to make more of it – just as Ukraine has sought a first-mover advantage against Russia, despite producing more. “I don’t think we’re going to beat China,” Thornton said. “The US continues to perform well time and again compared to China, which focuses on creativity and productisation,” he said.

Thornton argues – as do other defence tech startups –

that the real hurdle isn’t building different platforms – it’s the supply chains that support them. “The hard part is actually getting the stuff into the building,” he said: jet engines, solid rocket motors, and radar. Mack built and ran two jet engines in about eight months, a process he says traditionally takes four years; it acquired Exquadrum, a 24-year-old solid rocket motor company, in May for $50 million, beating out about eight other bidders by its account. Now almost half of Mac’s revenue comes from selling not only vehicles but also parts.

Mac’s approach is entirely different from some of his peers. Shield AI, founded in 2015, spent years as essentially a product company around its V-BAT drone before unveiling a second platform, the autonomous X-BAT Fighter, last October — and even that is being positioned as a big, deliberate bet, not a portfolio. Saronic, founded in 2022, builds only autonomous surface vessels with an integrated autonomy stack in hull sizes ranging from six feet to 180 feet.

Both have been rewarded for that discipline: Shield AI raised $2 billion this year at a $12.7 billion valuation; Saronic raised $1.75 billion to $9.25 billion.

The company’s strategy is very similar to that of Anduril, which is bigger, older, and a company against which every other defence tech startup is measured, objectively or not. Thornton himself makes the comparison, although he argues that there is a meaningful difference between the two companies. “Anduril’s playbook has been very top-down, starting with the software stack,” he said. “We’re very bottom up, starting with the hardware stack and then wrapping software around it.”

It’s a difference, yes, but Mach is still essentially operating in Anduril’s shadow. Anduril raised $5 billion in May at a valuation of $61 billion – more than 30 times that of MAC – and in March landed a 10-year, $20 billion army enterprise contract with a maximum cap, consolidating more than 120 separate procurement activities. Whatever Mac is building toward, Anduril got years and tens of billions of dollars in advance.

Thornton says that the field is not zero-sum.

He points to the scale of the problem: China reportedly makes about a thousand cruise missiles a day; The US builds one approximately every three days. “X company and Y company and Z company could all be manufacturing these things and it still wouldn’t be enough production,” he said. They also argue that the Pentagon would not structurally allow a monopoly, deliberately keeping two or three vendors alive in each category rather than choosing one winner.

Whether that’s a generous read of the competitive landscape or not, I told him that Anduril’s most famous co-founder, Palmer Luckey, has never publicly acknowledged Mac, as far as I can tell. Thornton dismissed any suggestion that Anduril has no interest in making room for Mach, telling me he respects Lucki and that they are “on the same team”, fighting for the same goal of Western sovereignty.

His investors, including Sequoia, Khosla Ventures and Ribbit Capital, are clearly indifferent. Founder-genius framing – Take away the Texas workshop and the MIT dropout story, including this one, and what’s left is a really interesting experiment led by a founder who, at least, seems to know what he doesn’t know.

Thornton has stated candidly that the hardest part of running a Mac changes every six months: first engineering, then sales, and now mass manufacturing, which he expects to dominate the next year. He says he tries to spend four or five hours a day thinking and “battling for the future,” sometimes taking colleagues off work to do the same with him – which, he admits, “can sometimes frustrate them.”

Regarding who puts pressure on him – who keeps a fast-rising founder honest – Thornton said the most valuable feedback doesn’t come from investors or even his executive team, who can end up in the same echo chamber as the CEO. “It comes from people who are actually doing the work,” he said.

He described regular company-wide forums, his COO’s idea, where employees get microphones and ask him anything. It started with Thornton quietly recruiting a few trusted aides to ask aggressive questions. Since then, it has become harder to control – and, he suggested, all the more useful for that. “I basically stand there for about an hour and have people in the company ask me the most invasive possible questions,” he said. It looks like he’s enjoying it.