Free with a sign-in.
The best 15 min of 42
What Nobody Tells You About Building Defense Tech Companies | Trae Stephens
Trae Stephens argues that defense tech founders should start with a focused product and build toward a broad, multi-domain company.
Minus One · Aditya Agarwal
- Trae Stephens is cofounder of Anduril Industries and Partner at Founders Fund.
The best minutes
- 23:45High valuations can make even successful startups poor investments6 min
- 30:28Anduril grew out of a search for software-led hardware4 min
- 34:20Anduril started with a deployable product, not its ultimate ambitions5 min
Also worth watching
- 39:18Defense startups must expand beyond a single product category3 min
- 12:01Palantir took a decade to reach $1 billion in revenue4 min
- 17:24Twelve companies generated over 95% of Founders Fund's returns6 min
- 2:50Palantir's culture won him over by turning ideas into action6 min
Skip the opening discussion of entrepreneurial quests and the Facebook office, hiring, and growth anecdotes.
Full analysis · 6 min read
Real Alpha · Nov 7
Trae Stephens says Anduril landed a program in 22 months
- Stephens says Anduril had a multi-hundred-million-dollar program of record 22 months after founding.
- That supports his case for sequencing, but not a general promise that defense sales move quickly.
- Founders should price financing against the next round and build only what field use requires.
Free with an account
Read the rest of the analysis
7 sections, the 3 other voices it tests Stephens against, and what it means for you. Every episode we cut gets one.
Check your inbox
We sent a link to your address. Click it to finish signing up. Not there in a minute? Look in spam.
or
Signed up before? The same buttons sign you in. The episode, its moments and the transcript stay free.
A quest can matter more than founder identity
Stephens says founders often reverse the order. They adopt an entrepreneurial identity, then search for a company idea.
He says stronger ventures begin when a problem becomes difficult to ignore, not when someone decides to become a founder.
That shifts the test from whether an idea looks like a startup to whether the founder will stay with it.
For example, Stephens was working in the U.S. intelligence community and trying to get Palantir's existing product into his agency. The agency told him to write a statement of work instead.
That resistance turned an abstract technology category into a problem he had encountered himself. Obsession may help a founder persist, but it does not prove a customer will pay.
“I don't think you find a good quest. I think a good quest finds you.”
Our read
The advice concerns the source of commitment, not a shortcut to product validation. A founder's genuine problem can sustain the work, but customers still have to confirm it matters.
Palantir paired fast execution with slow adoption
At Palantir, Stephens watched employees open their code editor, change the product, and launch it. That made execution visible.
It contrasted with the government meetings he described, where people focused on introductions, agendas, and commissioned white papers. His contrast is whether a team can turn discussion into a test.
Stephens moved from writing code to sales within weeks, after learning his engineering skills fit the role less well. His intelligence work and customer experience fit better there.
Still, Palantir took five years to reach $10 million in revenue and 10 years to reach $1 billion. He described failed approaches, including retired generals, consultants, and partnerships with large firms.
For founders selling to government, product iteration and customer adoption run on different clocks. Internal speed did not remove the work of changing customer behavior.
Our read
Fast internal iteration can coexist with slow customer adoption. Founders selling to government should budget for both, rather than treating product speed as proof of sales speed.
Fund returns make scale the investment filter
Stephens said Brian Singerman set him a first-year goal of 500 pitch meetings, without requiring investments. After about 200 meetings, Stephens says he understood the lesson. Many companies were not investable for that fund.
His own fund example is concentrated. Twelve companies generated more than 95 percent of its returns over 20 years.
On Stephens's illustration, 10% ownership of a $1 billion company yields $100 million, one-tenth of a $1 billion fund. That is why he asks whether an investment could become a $10 billion, $50 billion, or larger company.
Matthew Harris said on How I Invest in October 2026 that Bain invests $500 million annually to avoid chasing cycles. That is a different discipline about pacing, not evidence for Stephens's scale threshold.
“12 companies have returned 95 plus percent of all the returns to the fund.”
Our read
The power law explains this venture fund's bar, not a company's stand-alone quality or social value. Founders should identify whether an investor is underwriting a good business or fund-returning scale.
High marks can make success hard to finance
Stephens described companies coming in saying they were doing a $300 million round. The host, Aditya Agarwal, asks whether $300 million means capital raised or the valuation.
Agarwal separately offers a hypothetical pre-revenue, pre-incorporation round above $100 million. In his math, even a company at the $100 billion level might not return the fund after dilution and capital costs.
Stephens ties his pricing duty to Anduril's 10,000 employees and the next round. He says he must not raise at a price he cannot beat next time.
Agarwal said a high Dropbox private mark made later fundraising difficult when growth or the market blipped. He said taking a 20% haircut then could have given the company more flexibility.
Erik Bernhardsson said on 20VC in November 2026 that smaller, lower-priced rounds are often better when capital isn't scarce.
“we're doing a 300 million dollar round.”
Our read
This is not evidence that any lower price is safer. It makes the next financing case, rather than today's headline, the relevant test.
AI may support higher prices only with proof
Stephens leaves room for higher entry prices when technology opens new markets. He says AI may make ventures work that would not have worked a decade ago, using smaller teams and better margins.
But broad optimism about AI is not enough, he says. Investors must see a large market and a credible company.
Arda, Bob McGrew's manufacturing-automation company, is his example. Stephens said investors could pay a higher price because they see a large market and believe in the entrepreneur.
On How I Invest in November 2026, Jack Selby said AI can speed revenue and viable outcomes without giant exits. That describes potential faster viability, not proof of the margin or market assumptions Stephens requires.
Our read
Selby's point does not displace Stephens's pricing test. It adds a separate question, whether faster, cheaper execution can make a smaller exit worthwhile.
Anduril won access by starting with a fieldable product
Stephens says Anduril's seed deck listed fighter planes, submarines, and low-cost cruise missiles. Yet the team chose a product it could field quickly, rather than attempting its final ambition first.
That is the logic behind his SpaceX analogy. Falcon 1 came before Falcon 9, reusability, Starlink, and Starship.
Take Sentry, the counter-intrusion tower. It combined solar power, radar, infrared and optical cameras, edge computing, and communications. Stephens said the plan was to deploy it within six months, without a basic-science problem.
Instead, integration exposed a gap. Existing pan-tilt units could miss commanded angles, and repeated movements made the pointing error matter. Anduril built its own unit.
Stephens said the company started in June, had a test unit in September, and deployed its first towers in December. Within 22 months, he said, Anduril had a multi-hundred-million-dollar government program of record.
He said the program was still running and worth billions, with the tower carrying the portfolio's best margin profile. Operational access and real-world data then helped Anduril build software for its wider products.
“SpaceX did not build Starship from day one.”
Our read
The transferable move is to choose a fieldable product that earns customer access and feedback. It is not a promise that hardware will be easy or sales cycles short.
Defense prime ambitions require crossing categories
Stephens says a company aiming to become a defense prime cannot stop at one product category. He points to Boeing, Lockheed Martin, Northrop Grumman, General Dynamics, and RTX, which build many products.
That makes the wedge a starting point, not the company's full scale story. A narrow first product does not establish the capabilities needed for other domains.
Stephens says Anduril is building a shipyard for Virginia-class submarine supermodules. He says it has to work with General Dynamics Electric Boat to make the project happen.
That describes a partner relationship alongside competition with the established primes, not free territory. His model raises a hard operating question for a new entrant. Can it extend across domains without losing the speed and product discipline behind its first deployment?
Our read
A narrow first product can open a door, but prime scale requires adjacent capabilities and relationships. Treat expansion as a long-term operating commitment, not a slogan attached to a first contract.
What it means for you
- Map the next round. Model dilution, future capital needs, and a plausible next valuation before accepting a headline price. Write down which assumptions could make that next round difficult.
- Set a field test. Choose a product milestone that can reach real users quickly. Track deployment time, customer access, and what field use teaches about the software.
- Audit component gaps. List the hardware your product depends on and test it under real operating conditions. Build a component only when existing options fail the required job.
- Name the expansion path. If your ambition requires a broad company, identify which neighboring products or domains could follow the first wedge. Test whether customers and partners support that path.
What to watch
- Whether Anduril can preserve its product economics as it enters more defense categories.
- Whether AI-enabled smaller teams produce margins that justify higher startup entry prices.
- Whether founders plan financing around the next round rather than today's headline valuation.
Send this to founders raising hardware rounds and CEOs expanding into regulated markets.
Watch the minute Stephens gives Anduril's 22-month program timeline
Written by Real Alpha from the episode's transcript and the other moments it links, then checked against the transcript. Quotes are the speaker's exact words.
Minute by minute
- 0:26A good quest finds the founder
Stephens says the best companies often begin when someone becomes obsessed with a problem that will not leave their mind. He argues that starting a company just because someone identifies as an entrepreneur can lead to bad ideas.
- 2:50Palantir emerged from a real need
Stephens says he was working in the U.S. intelligence community and had no plan to enter tech when he encountered Palantir. After his agency resisted using the existing product, he contacted the small team, visited Palo Alto and decided its people were a fit.
- 5:28Execution revealed the right people
Stephens says he could often tell quickly whether a founder or company was a fit, though he tried to respond respectfully when passing. At Palantir, he saw colleagues build and demonstrate their work rather than spend meetings planning around bureaucracy. He says the company also let people shift roles to match their strengths.
- 11:42Palantir's growth took years
Stephens says Palantir took five years to reach its first $10 million in revenue and 10 years to reach $1 billion. After six difficult years and unsuccessful efforts including partnerships with large firms, Peter Thiel asked him to join Founders Fund; Stephens says it took nine months before he joined.
- 17:24Investing took hundreds of meetings
Stephens says a Founders Fund partner challenged him to take 500 pitch meetings in his first year, without making investments. The experience showed him how few companies were investable and helped him recognize teams and opportunities with the potential to return a fund. He says repeated practice also helped the partnership align on what makes a strong investment.
- 21:19A few companies drive returns
Stephens says 12 companies returned more than 95% of Founders Fund’s returns, including Facebook, SpaceX, Airbnb, Spotify and Stripe. He argues that investors need to assess whether a company could become a $60 billion to $100 billion business, since a smaller win may not matter much for a billion-dollar fund.
- 22:55High valuations raise the return bar
Stephens says large AI rounds can require a company to reach a trillion-dollar outcome just to deliver the venture returns investors once expected from a more ordinary exit. He argues that investors need to stay disciplined and founders should consider dilution, future fundraising and downturns before accepting a high valuation.
- 29:11AI can justify selective price increases
Stephens says AI creates markets that may not have worked a decade ago and can let companies operate with smaller teams and better margins. He says that can justify somewhat higher entry prices for companies pursuing genuinely large opportunities, but optimism about AI alone is not enough.
- 30:01Anduril began with a thesis
Stephens says he reviewed public federal-contract data and met with hundreds of companies before concluding that defense needed a software-driven, hardware-enabled business focused on low-cost autonomous systems. Peter Thiel urged him to build it, and Stephens recruited Palmer Luckey, who later agreed to join after leaving Meta.
- 34:11Start small, then expand the product
Stephens says Anduril took an iterative approach, like SpaceX moving from Falcon 1 toward Starship, and began with Sentry, a counter-intrusion tower. The team deployed a test unit in September and its first towers in December, then secured a multi-hundred-million-dollar government program 22 months after founding. He says Sentry helped the company build and improve the software brain used across its products.
- 39:18Defense has no free space
Stephens argues that a company aiming to become a defense prime cannot remain in a single niche: Anduril competes with major contractors and sometimes works with them. He points to the company’s work with General Dynamics Electric Boat on submarine modules as an example of how its activities expanded beyond its original plans.
Published Nov 5, 2026.
Transcript · 37 min read
Minus One, hosted by Aditya Agarwal, with Trae Stephens.
Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.
I don't think you find a good quest. I think a good quest finds you. People are really desperate for like the, you know, the insight. How do you know that this company idea is going to be a great company idea? And the reality is, is like, it's actually really obvious most of the time. Almost all these companies are completely uninvestable. If you believe that you're operating in free space, it's just not true. There is no free space left in defense. And this is actually an important lesson for entrepreneurs. SpaceX did not build Starship from day one. One of the things I admire in your career is that you seem to have made very good decisions or very interesting decisions about things you have picked to work on, either people or companies.
What's your organizing framework for how to think about good quests and ideas worth working on? I don't think you find a good quest. I think a good quest finds you. I'm a huge Monty Python fan. I don't know if any of you are. But there's this scene in Monty Python and the Holy Grail where King Arthur and his knights are just randomly wandering around and he's acquiring some of them. And then God appears in the clouds and they all start groveling. And he says, King Arthur, I have a quest for you. And it's a quest for the Holy Grail. And that's like what ends up happening through the whole movie.
But King Arthur wasn't going around knocking on doors being like, what are your problems? Tell me about how I might be able to build enterprise sass. No, it's just God gave him the quest. And I think too often in entrepreneurship, people, their identity is I'm an entrepreneur and now I have to find a quest. And I think actually the best companies are I'm not an entrepreneur. I'm just a guy. And I like I'm obsessed. I become obsessed with this thing that can it won't leave my head. And that's that's the beginning of the entrepreneurial journey, not the other way around. And I think getting that mixed up, it leads to a lot of bad quests, really dumb companies.
Read the whole transcript
You know, I love that answer. I mean, I actually think that this is something that Zuck used to talk a lot about at Facebook, which he was that the like, you know, doing a company is really hard. Most companies don't work. It is absolutely the hardest thing to kind of do in terms of building a career. So you should almost do a company as an act of last resort. if this is the best vehicle that you can find to bring an idea to life, then you should kind of embark on a company. And in some ways, it should kind of, you know, really draw you to it as opposed to the other way around. So maybe with that framing, Trey, how do you recognize when you are getting pulled, particularly in a super noisy environment like we live in today, where there are so many different distractions and ideas and people, and a lot of it sounds pretty enticing. And frankly, I'm assuming that the time when you joined Palantir, it was a bit easier because it was not as noisy then, right? So how do you almost winnow out the noise that invariably
“And after spending two and a half years working in the government, I believed nothing more than that the bureaucracy needed to be broken.”
Actually, I didn't realize that you were at Palentry in 2007. I have an interesting story, which is that in that time period, Facebook was scattered through a bunch of different offices in downtown Palo Alto. And we would keep on like getting new office space and then kind of growing out of it. But then I got introduced to this guy from Palantir, Aki. And so Aki's strategy was basically like, I don't need to look for office space. I'm just going to find somebody from Facebook. And as they keep on vacating office space, I will keep on taking it. And so Aki and I would meet every three months.
He's like, hey, do you have any office space for me? I'm like, yeah, this is the one we're going to be leaving soon. And he would just end up taking, he's a great guy. Well, you know, the beautiful thing about that story actually is that Aki was the CTO at the time. And I think that the reason this is relevant in like the context of hiring is that when you're building a startup, you have this idea, I think, that you should hire people for the specific skills or whatever. And the reality is, is like no one at Palantir ended up doing the thing that they were hired to do. Like Aki did a lot of things.
Like, did he write code? Yeah, sure. Everyone was writing code. You know, I started writing code at Palantir too. And then in the first three weeks, I was dinking around with like a training exercise, onboarding, during onboarding. It's funny how like these people have come around to do crazy stuff. So Bob McGrew was the head of engineering. Yeah, Bob was here just a few months ago. Yeah, Bob is, he was the head of research at OpenAI. And then he recently just started a company called Arda that's doing manufacturing automation. I think they just kind of did a huge public thing today. Yeah, yesterday.
They just launched publicly actually on Monday. But yeah, this week. But Bob, who I'd never met before, walked up behind me and he was like, hi, I'm Bob. Who are you? And I was like, Trey. And he's like, never touch my code base ever again. Your code is garbage. And I just kind of laughed and, you know, within a week I was running sales, which is exactly what I should have been doing, actually, because, you know, a top 5% engineer working in the intelligence community is like a bottom 5% engineer working on a software company in Silicon Valley. So I think it's like that having that flexibility and awareness of how skills align was like a driving force at Palantir that worked for us really well.
One thing based on kind of the slight tangent that you took earlier that I want to mention, I think this is really important advice for founders in this room, which is the longer your diligence process drags on with a venture capital firm, the less likely they are to invest. right i think that oftentimes founders think that it is an indication of a lot of interest when people take the fifth or the sixth meeting it's actually a case of like they are struggling to find a reason to kind of invest because they generally know in the beginning they generally know pretty quickly whether they love it or not so anyways coming back to uh to your stories right here um okay so you were at palantir for six years six years that's right and then you make your way over to founders fund what made you transition kind of like from being an operator to kind of an investor they just kind of like They wanted to work with Peter Thiel.
Was there some underlying motivation or kind of thought process there?
“it took us five years to get our first 10 million in revenue. It took us 10 years to get to a billion.”
I want to get back to what the first four or five years at Founders Fund were like before you also did Entral. But coming back to your point about Zuck, you are right that Facebook led a charmed life in the beginning. However, internally, Zuck always kept on raising the bar. We never felt like we were winning. It was always like, why are we not winning more?
Right. Like why? Like every like there would be either some fucking competitor that we needed to crush or some metric that like was the next. Like if you can get to 100 million users, why can't we go a billion? If it's a million miles, if it's a billion miles, why can't it be a billion dows, right? And so it was always a case of our ambitions grew faster than I think our success. And so my almost like visceral memory of Facebook is always like, we can do more. We must do more because it's almost, it would be a shame to not capitalize on this opportunity that we have because these opportunities don't come around very often.
So I think Mark is just like, I mean, Mark is still pushing the company super hard today. And I think this is one of his best qualities. You come to Founders Fund.
“12 companies have returned 95 plus percent of all the returns to the fund.”
Okay, so this is a good segue, which is perhaps, I think that there was broad consensus around outcome sizes and kind of market segments for those outcome sizes, I think across our industry, broadly, you know, maybe up until three or four years ago. But I think over the last three or four years, it feels as though a lot of kind of like, you know, the kind of, you know, like, hey, this could be a $10 billion company. Now it's like maybe like, you know, a much bigger kind of like, you know, potential size than we all thought it was going to be.
So I guess maybe the question here is that how have either you yourself at Andrel or maybe at Founders Fund started to adjust some of your own priors on how big some of these outcomes can be in a world where we actually have kind of the tailwinds of massive acceleration through AI?
“you have to be responsible in the way that you raise because otherwise you're literally playing a game of chicken with your company”
Yeah. Maybe take us through the founding of Andrel. Like, you know, it sounds like I don't know if the story is accurate, but it sounds like you kind of pitch, you know, Palmer on this and you kind of got him excited. And the two of you and along with a bunch of other folks started the company. But did you just, you know, you're like, OK, I'm an investor, but I also can start companies on the side. I know Founders Fund has done this historically, but did it feel pretty natural to kind of like follow that kind of like calling or that kind of question in your head?
“I thought I had the ace in the hole. I was like, well, you know, Peter, Palantir and SpaceX were both founded by billionaires.”
Some stories are stranger than fiction. That's true. So what does it mean to, I mean, I get the soundbite, I understand it, a software-defined
“The company was started in June. We had a test unit in September. We had first towers deployed in December.”
We could only do that because we had operational access, the customer trusted us, and the And we were able to continue iterating to make sure that it did exactly what we wanted it to do.
“There is no free space left in defense.”
authenticity and kind of like relaying us with a bunch of your stories you're always welcome back at svc thank you so much thanks guys appreciate it
