• D.C.
  • BXL
  • Lagos
  • Dubai
  • Beijing
  • SG
rotating globe
  • D.C.
  • BXL
  • Lagos
Semafor Logo
  • Dubai
  • Beijing
  • SG

Updated Jan 3, 2024, 12:56pm EST

Defense tech is having its moment in Silicon Valley

U.S. Defense Department
Title icon

The Scene

After Silicon Valley shunned ties with the U.S. military for years, defense tech is having a moment in the industry.

For the past few years, more startups and investors have been building and funding companies targeting the Pentagon as a customer.

The current trend contrasts with decades of going their own ways, with the U.S. military getting stuck in a procurement loop, buying outdated technology from entrenched contractors. At the same time, the tech industry had companies like Apple taking American technological knowhow to China in exchange for cheap labor and robust supply chains.


Now, as U.S.-China tensions rise and Russia’s war in Ukraine drags on, what’s emerged lately is a kind of techno-patriotism in Silicon Valley, combined with a belief that a byzantine procurement process is keeping the best private-sector technology out of the hands of American military personnel. The AI craze has supercharged that environment.

Dan Gwak, managing partner of Point72 Private Investments, has seen both sides of this equation. In 2008, he left his job in finance to join the U.S. Marines, where he earned a Purple Heart in Afghanistan. And he’s also worked for In-Q-Tel, the CIA-backed venture firm created to tap private-sector tech to aid intelligence goals.

Now, he’s investing in defense tech and helping startups navigate the government-contracting maze. I spoke with Gwak about the struggle to get the latest tech into the hands of soldiers and what Silicon Valley is doing right and wrong on that front. Read below for an edited interview.

Title icon

The View From Dan Gwak

Q: What are the hardest problems to solve in defense tech?

A: AI, especially autonomy. Outside of Shield AI, it’s pretty hard to find a truly successful company like that. A lot of commercial autonomy startups are struggling. That is truly difficult technology. I have full confidence the defense primes are not going to crack that. That’s why we are willing to fund a startup to do that. If we can get autonomy right, that’s absolutely going to be game changing in a near-peer conflict. And it’s something that, if you get right, the government will basically have no choice but to buy. Air, land, and sea autonomy are all areas that are super important.

Another area is large language models. Great power competition is much more likely to be like a Cold War, and then it’s about decisive technologies that deter adversaries’ actions. What really matters in that scenario is intelligence. What is the adversary thinking? What are they currently publishing about? What can we glean from what people in that country are saying. Even within our own country, sometimes it’s very hard to figure out fact from fiction because there’s such a prevalence of misinformation. But misinformation is going to be a huge part of future conflicts.


What LLMs can do in creating a whole bunch more, or identifying a whole bunch more, information campaigns is truly profound. It’s not just like LLMs can save you the writing of the thing. They’re starting to actually be able to log into SaaS products like Twitter, and go through the workflows themselves. It’s like a 100x, 1,000x productivity increase for human beings. These cutting edge technology waves that are the big game changers, the primes aren’t well positioned to address.

Q: When you talk about autonomy, that’s fairly broad. Do you have examples?

A: The hard part is actually the autonomy software. That’s the control system that can take in the sensor data, and make sense of it to control the vehicle. I can imagine a world where there’s a software adoption pathway, where software ends up going into the existing platforms. You’re probably not going to build a whole new nuclear submarine just for the autonomy capability. So that software adoption pathway is important. You’re starting to see that in, for example, grounded autonomy. There’s a program with robotic combat vehicles [RCVs]. Primes compete for the hardware component of it and then they’re having a separate acquisition process for the control software part of it.

Those things are separated purposefully because the defense primes have a lot more experience building that tank body or whatever. And the startups have a lot more experience building the autonomy software.

But the biggest impact of autonomy isn’t just taking the existing platforms, and then making them autonomous. What’s truly game changing is things you can do that you previously couldn’t. An example of that is the Marine Corps NMESIS program. They have an idea of creating an autonomous vehicle that can launch missiles, launch artillery, and then autonomously get out of there.

Why is that so important? My father is actually an artilleryman. When you launch artillery, it goes through the air, gets caught on radar, and they know where that’s coming from right away. They’re sending artillery back instantly. Oftentimes, that means an artilleryman can get one round and then they’ve got to pack up and get out of there as soon as they can. If we can autonomously do that, it’s 100x better.

Q: Do you need lobbyists to sell to the government?

A: The fundamental reason that it is so hard to sell to the government is what I call the Triangle of Death. The users, the budget, and the authority all live in different places in [the Defense Department]. To make matters worse, they don’t know about each other. That’s the crazy thing. The user is the uniformed military service person who is facing off against the enemy, maybe a drone pilot or an intelligence officer who needs to understand what’s being said in the Chinese Twittersphere.

That user usually loves talking to tech companies because they can tell right away what’s going to be massively game changing and really help them in their job. But if you ask that user how they would go about purchasing something, they have no idea.

When I was a Marine, I got issued a rifle. I could not tell you who made the decision that was the rifle we were going to carry. So for startups, the users are usually fairly easy to identify and get to, but identifying where that budget and where the authority lives are the really hard part about defense tech sales into government.

Q: So do lobbyists navigate that?

A: Every year, there’s the National Defense Authorization Act, which contains the $800 billion of defense funding where the budget all starts and flows from. In order for a program to get funded, it has to be approved by Congress. And that’s where lobbyists come in.

The major weapons platforms we have today that cost hundreds of millions, billions of dollars, those are all programs of record. The F-35, for example, is a program of record passed in the NDAA for funding. A defense tech company has to scale to a venture size exit by generating revenue purely from DOD. That’s only going to happen if they can actually get new programs funded.

There are two ways to look at it. One way would be to say, ‘That’s such a shame that it’s so hard to sell into the government that it requires lobbyists.’

Another way to look at it, which is a bit more pragmatic and healthy in some ways, is to say, ‘That is the system.’ It’s not like we hire lobbyists but portfolio companies do. Our best companies see that as a part of how to win. Some folks might say, ‘I’m just a two-person startup. How am I supposed to have the budget to hire lobbyists?’

We live in a capitalist system and there has never been greater interest in defense tech investing than there is now for a qualified team of great technologists that have experience and understand user problems. If you can’t raise financing in this kind of environment, and therefore can’t hire lobbyists, that’s probably the system working the way it should.

Point72/Edited with Adobe Generative Expand

Q: A lot of people in Silicon Valley think the government is using old technology from the old guard of defense contractors. And they should get rid of that and instead, buy new technology that’s better, or maybe lower priced. So that would require less budget.

A: That’s ultimately the burden of technologists, to bring something better, or cheaper, or both. The job of a defense tech startup is to make that obvious. That’s where the conversation often breaks down. The very early stage startup might go to a user, like the Space Force. They’re thinking about spending a billion dollars on a satellite program. The startup says, ‘I can do it for a lot less.’ Is that a risk that the government should take on?

The system is working well because it should be the venture capitalists vetting the technology. They fund that team to develop a product and provide proof points where it becomes more obvious to the government. The government shouldn’t have to take this outsized risk with just a team on a piece of paper saying that they can do things right.

That’s also what allows startups to charge the government differently than defense primes and make better margins that are much more like technology margins. The defense primes have a business model of cost plus. The government says, ‘We want to come up with this kind of airplane, and we’ll pay you to develop that and we’ll pay you some profit on top of your cost.’ When it comes time to put them into production, they’re going to cap the profitability of the defense prime.

Defense tech startups are different. You don’t need to pay for us to develop the product. We’ll raise the capital, we’ll develop a product, we’ll bring it to you, it will be better, and we will sell it to you at the price that we want to sell it. That’s a net win for the taxpayer and that’s fundamentally what you want to happen. It puts the risk in the right places.

Q: So it’s possible to push the incumbents out and the best technology wins in the end?

A: Shield AI is a great example of that. They’re growing very fast. Some of that is new demand and a lot of that is taking over existing demand. That’s not to say it’s easy. I go back to the Triangle of Death. They had to solve that problem. A defense tech company like Shield AI has to find the user, get them enthusiastic, find the budget, find the authority, and stitch all those things together into a deal.

The Triangle of Death is still a problem. The biggest injustice is to the user, who is risking their life to fight this nation’s battles. Yet, they have so little say in the tools that can help save their lives because they have no access to the authority and the budget that can fund those tools. Those decisions are getting made in a building they’ve never heard of by people that they’ve never met.

Q: The message from the military now is things are changing. We’re trying to engage more. What does that actually mean?

A: I would hardly say that the problem is solved. A lot of the government efforts are focused on helping startups find the users, but that’s not the hard part. And any startup that comes to the government and says, ‘We want our customers brought to us on a silver platter’ is already wrong. Collapsing the budget and the authority is the hard part. I would encourage DOD to spend their effort making that part easier.

Q: So you think that the budget authority should be closer to the end user?

A: I couldn’t have phrased it better. And the more we can make that happen, the healthier the whole system will be. This whole bureaucracy of DOD acquiring technology was stood up at a time when the most needle-moving technologies were things like large aircraft carriers. I am not saying that we should ask some lance corporal what nuclear submarine we should buy. The budget was approved by Congress, because that was a national level priority.

Today, it’s about the edge, it’s what’s forward deployed. When we think about America’s battles moving forward, all these services are talking about a much more distributed warfighting force. In a great power competition, you don’t want to have large concentrated forces sitting in one place. In the kind of island hopping campaigns that people are thinking of in the South China Sea, forces are naturally going to be distributed.

If that is the fight of the future, we need to push technology to the edge. Which drone is useful is absolutely something that somebody who is on the ground, carrying a rifle, can make a better decision on than the bureaucrats sitting back in the Pentagon. Change is necessary to keep up with that.

Semafor Logo