Rohan Goswami:
Welcome to Compound Interest from Semafor Business. I’m Rohan Goswami, Semafor’s business reporter, and I’m joined as always by Liz Hoffman, our business and finance editor.
Where are you joining us from today, Liz?

Liz Hoffman:
Hi, Rohan. I’m in Semafor’s offices in downtown Washington, DC, down here this week. It’s a great backdrop for the conversation we’re going to have today. We’re talking to David Ulevitch, who’s a partner at Andreessen Horowitz, also known as A16Z, the big venture capital firm. He runs their American Dynamism Fund, which is kind of the America first fund inside the OG America first firm. A16Z was kind of ground zero for this ideological shift in Silicon Valley. Their founder, Mark Andreessen, was an early convert to Donald Trump’s second term campaign, and agenda, and has sort of epitomized this investing shift towards; national security, defense, space, robotics, energy, minerals.

Rohan Goswami:
Real things.

Liz Hoffman:
Real things, but that all have this slightly nationalist, slightly hawkish and slightly right coded undertone.

Rohan Goswami:
It is also interesting culturally at A16Z, it is a remarkable shift. I mean, Mark Andreessen, the ‘Software Is Eating The World’ essay was iconic. Now you’re watching them sort of realign and try and find a new blueprint in a world which has been totally scrambled and upended by AI.

Liz Hoffman:
That’s why I really wanted to have David on the show. The entire investing skillset and ecosystem of Silicon Valley was sort of oriented around really capital light, infinitely scalable code. They sort of very proudly built light stuff. We had an Uber and then we got an Uber for everything else. They were backing these business models that took very little physical risk, basically never touched hardware, almost never touched people, and were kind of marketplaces connecting tailors with people who needed their pants hemmed and taking a vague there. They very proudly built things that didn’t weigh very much. That’s sort of the simplest way to see the transformation in venture is that they’re building really heavy, really expensive stuff. The question is, A, why?

Rohan Goswami:
That is always the question.

Liz Hoffman:
As always the question. But how much of that is that they just like this stuff? Just ideologically, they have shifted towards a more conservative, more America first thing. They want to put their money there regardless of whether they have a real right to win. And then B, are there going to be any good at it? Huge parts of the way the US, and this is happening to some degree globally, build national security is now in the hands of people who haven’t done it before.

Rohan Goswami:
David is absolutely the right person to talk to about all this. So let’s take a quick break and we’ll come right back with him.

Liz Hoffman:
David, welcome to the show.

David Ulevitch:
Thank you for having me. Good to be here.

Liz Hoffman:
So just to start, what is A16Z’s American Dynamism Fund, and what is the origin story there?

David Ulevitch:
Okay. You want the long answer or the short answer?

Liz Hoffman:
We want the long answer. If it’s too long, we’ll cut it down.

Rohan Goswami:
Always. We want the long answer always.

David Ulevitch:
Okay. The long answer. The American Dynamism Fund is a fund that we have here at Andreessen Horowitz that is focused on investing in companies that we say serve the national interest. That’s a broad term, which largely means investing in defense, aerospace, space, broadly, energy, the supply chain, public safety, manufacturing, all these things that contribute to the foundations of how we live and work that promote national security, that promote economic security and resilience, and sort of make this the place that we all are excited to either be from, come to, build our businesses in, live in, and make the American dream more real.
On top of that is a broader sort of umbrella that encompasses lots of things around education and healthcare, things that we may not invest in, but we will often talk about because they’re important. The other thing that I think is distinctive for American Dynamism is, not just the focus on the categories of defense and energy, public safety, et cetera, but often these are companies that sit at the intersection of very, very advanced software, as well as hardware. It’s where atoms meets bits, which historically has not been, I would say, en vogue for the venture capital industry, certainly over the last 25 years, although now it’s becoming more popular, I think probably due to us.
So that’s largely what American Dynamism is. And then I’m happy to tell you about the origin story about how it came about, both as a firm practice, as well as for myself.

Liz Hoffman:
Yeah, please.

David Ulevitch:
I ran a cybersecurity company called OpenDNS for a long time. I was never a veteran. I never served our country. I didn’t even know that many veterans, but then my company got bought by Cisco. Two things happened. One was that I got to work with a lot of veterans. Cisco hires a lot of people that served our country as they come out of the service. Then the other thing that happened at Cisco is I became an executive sponsor, which is kind of like a glorified escalation point inside the company for a bunch of customers that were parts of federal government, so people that were part of the intelligence community, people that were part of the Pentagon. If they would go out and buy 50 firewalls and would be told that it might take four months or six months for delivery and they didn’t like that, they would call me and escalate.
I ended up getting to spend a lot of time with those customers and really grew to realize and learn that the people that worked for our country in the armed forces, or just in the federal government, or in the intelligence community, these were people that were just super apolitical. They were very patriotic. They deserved the best technology and capabilities. They often did not have them and they just wanted to make sure that the American experiment would continue and that we would remain the world’s superpower. They were just very, very patriotic in their mission.
When I left Cisco, I had met Mark and Ben. Ben made the mistake of saying if I joined Andreessen Horowitz I could do whatever I wanted. A year after I joined, I really was trying to think about how I wanted to spend my time. My background was in enterprise software, but I was much more excited about doing something that I thought would have very long-term meaningful impact. Mark introduced me to Palmer Lucky, the founder of Anduril. I co-led the Series B of Anduril with a woman named Catherine Boyle, who was at another firm at the time. That’s how we met, so we were rivals. I think she probably didn’t even like me because she thought she was going to do the deal herself.
Then fast-forward, she and I got to know each other better, we talked about deals and about two years later toward the end of 2021, I ended up recruiting her and said, “Hey. Look, I believe in this opportunity that we have to invest in introducing these fundamental technology shifts into these categories like defense, and public safety, and space, and energy. Let’s make a full practice out of this.” She was putting to paper a lot of the things I believed in, and so I hired her at the end of 2021. We launched the American Dynamism practice in 2022 and then raised a fund, and now we’re on our second fund and off to the races.

Liz Hoffman:
I’m curious, American Dynamism, pretty neutral and all the things you’re talking about, national security, energy, in theory should be bipartisan or apolitical. I’m in Washington this week, and so maybe I’m just simmering in the partisan stew, but what do you make of the fact that a lot of these areas feel right coded these days? Is that a problem for you, as you think about fundraising or investing?

David Ulevitch:
Yeah. I think one of the things that I’ve discovered, our firm is very active in politics and policy, so I get to see this stuff firsthand is that most of these issues are actually very bipartisan. They may not always feel bipartisan, in the talking points. When you really spend time one-on-one with policymakers, they’re very bipartisan. I think they feel right coded because we have an administration that loves to lean into the partisanship and the rhetoric. Frankly, we have, I think, Democrats on the left who love to lean into their view of that same conversation, the other side of the coin. When it comes to getting work done, there’s a reason that a lot of these critical committees are … I mean, they’re all by definition, bipartisan.
What’s right coded to me feels like the momentum and desire for change, and the appetite to probably break a little bit of glass by making change. That probably does feel maybe more right coded, because we have an administration that, again, leans very heavily into the, “We don’t care about the way they used to do it. We don’t care about what was done in the past. We have to do things differently.”
I actually think that, hopefully, will continue no matter what the next administration of leadership looks like because I think that people recognize that we have a system of incumbents who are very slow moving, and stodgy, and not using the newest and most advanced technology. We have production and manufacturing capabilities that have really slowed down and atrophied, and been outsourced, and put too much supply chain resilience on people that may not be able to supply us in the future or may even be adversaries. I think that the right coded piece is really just the appetite for rapid, radical change.
I think the core issue is energy abundance. We have an insatiable thirst for energy. It doesn’t matter if it’s for AI data centers, or for electric vehicles, or for national security and grid resilience. We just have this insatiable thirst for energy. That’s a fairly bipartisan view.

Liz Hoffman:
Yeah. You wrote last August, “We believe investors, not government, hold the key to defending America’s future.” Pretty bold statement. What did you mean by that?

David Ulevitch:
It’s not appropriate for the government to endlessly fund advanced R&D that may not have an ROI. We will take on that risk as venture capitalists, where we think we can find the best entrepreneurs, we think we can motivate the best founders to apply the best technology to manufacture things, and then just sell them to the government. I think people don’t always realize that that’s not the traditional way that the government has been doing things like defense R&D broadly for the last 70 years. What normally happens is where the government will say, “We need this capability and we will give hundreds of millions of dollars to a major prime contractor to work on this for a while.”
Then the contractor will take much longer to do it, and it will go way over budget, but the profit margin’s already baked in for that prime contractor, the incumbent, so it doesn’t matter that it’s over budget, it doesn’t matter that it took too long because if something was supposed to cost 100 million, but then it costs 200 million, but your profit margin guaranteed to be 10%, well, you actually just doubled your profit margin by going over budget.
And look, for things like aircraft carriers, maybe that’s appropriate because it’s a very exquisite platform, but when it comes to things where we want things that are low cost, high volume, high rate of production, like drones, there actually is a better way. A commercial way is better, where we fund the private capital to it and then the government will just buy it if it meets their capability specs.

Rohan Goswami:
Let’s talk about drones actually, because I did see some crazy stat about how basically asymmetric, the weapons that we’re using in Iran, specifically, the Shahed drones are. These $100 million weapons platforms that are very limited quantities, against these really cheap, lightweight drones that the Iranians are fueling.
You’ve talked about the process, and the process from your eyes and a lot of people’s eyes is broken. What should we actually be building? What are we not building in the US that you want to invest in more of, that you are invested in, that you think is really critical?

David Ulevitch:
Yeah. Well, I mean, I think advanced interception capability that’s at a high rate of production and at a low cost is a perfect example. We have phenomenal, phenomenal weapons capabilities. We have things like the Patriot missile, but you cannot take a Patriot missile, which could cost anywhere from five to $8 million per shot. Not only do we have limited quantity, but just the asymmetric economic cost is incredible, to take something down that costs a quarter of a million dollars, and of which there’s essentially an unlimited supply of the Shahed drone. They keep making more of them. Russia now makes essentially an equivalent Shahed drone.
You need to have something that’s both economically viable where you’re not losing the economic asymmetry, but you also need to be able to do high rates of production. That’s something that the primes have not demonstrated any ability to do, or appetite to do, despite multiple requests, I think, from the government to speed up production, to open up new factories. I think the demand signal from the government is now very, very clear. There are startups that say, “Hey, wait a minute, we understand the Elon philosophy of the factory is the product. We actually can focus on advanced automation, applying robotics to these processes, to verticalizing more of our supply chain and manufacturing capabilities. We can actually switch out some of the components for lower cost components. We can actually use software for the guidance systems, and the command and control systems, and not have to use very, very exquisite laser systems or optical systems.”
That’s happening now. I think we’re in the very early innings of seeing this on the battlefield in Iran. We’re seeing now that low cost drones are being used, but I think that this is probably the last major military engagement where we will not see these new startups have a much, much larger position on the battlefield than they have today. There’s pockets of it in today’s battlefield, but I think this is the last fight where it’ll be largely incumbent driven.

Liz Hoffman:
The other big storyline in defense over the last couple weeks is Anthropic. Let’s talk about the Anthropic of it all. The company’s in the middle of a huge fight with the Pentagon. OpenAI came in and took the contract. Sam Altman seems to have some regrets about how that all worked out too. Did the calculus of doing business with the US military just change?

David Ulevitch:
I think that, from my vantage point as a venture capitalist, I’m in New York, but ostensibly in Silicon Valley, or at least in the Silicon Valley [inaudible 00:13:28].

Liz Hoffman:
They let you keep VC card when you move to New York City?

David Ulevitch:
They do. Yeah.

Liz Hoffman:
That’s nice of them.

David Ulevitch:
Yeah. I think my takeaway actually, and I was talking about this with one of my partners the other day, I actually think a lot of this has to do with talent, and it’s almost like it’s a sorting algorithm where the people that sort of … If you think back to when Google employees protested Project Maven, that became a sorting algorithm for Google where a lot of people said, “Hey. I want to work at a place that refuses to make sure that our men and women who serve this country have the best capabilities.” They don’t want to work at that company, or they do want to work at that company, and it kind of sorts people out.
I think with the Anthropic, OpenAI decisions to support or not support working with the government and letting the government decide, it’s kind of that same sorting algorithm. There’s going to be employees who leave OpenAI and say that they don’t want to be a part of OpenAI, for whatever reason. They don’t agree with OpenAI’s decision to work with the government, or they believe that there should be extrajudicial process and CEOs of companies should be able to overrule the government, and they want to go work at Anthropic. They kind of like the God complex, CEO mentality and they want to be part of that. I think there’s other people that say, “Well, we elect leaders, the leaders make laws, there’s lots of oversight, there’s ways to provide judicial oversight and congressional oversight. I trust that democratic system, but I also want the people that serve our country to have the best capabilities.” Those people will stay at OpenAI.

Liz Hoffman:
It’s interesting because Anthropic, there was some self-sorting that happened on the way into the founding of these companies.

David Ulevitch:
Totally. This is very consistent for Anthropic.

Liz Hoffman:
Yeah. It’s one thing for the Pentagon to say, “Look, we don’t like Anthropic. We are mad at Dario,” and cancel the contract. That is a thing that happens all the time, and should in military contracting. It’s another to try to put a private company out of business, which, depending on who you ask, is what would happen if no company that did any kind of business with the Pentagon was allowed to use Anthropic in any of its systems.

David Ulevitch:
That seems both highly unlikely and also not exactly what’s happening. I think that people that have watched this administration know that they sometimes say the extreme form of something, and then they generally do something that is not nearly as extreme as what their original sort of headline rhetoric is.
I think Anthropic is doing incredibly well. I think it’s a very impressive company. I think it’s their decision not to work with the government. I think that will have, like any decision you make in business, will have consequences. I don’t think it will be a company-killing event. I think the company is a generational company. I think that there’s plenty of other companies that say, “Wait a minute, not only do we want to work with the government and we’re fine with it, but we don’t want to put ourselves in the position of trying to make extra judicial decisions about how technology is used by these customers who have far more information, and intelligence, and insight into the use cases and the situation.”
I can’t even personally imagine ever putting myself in that situation. Not just the legal liability it would open me up to, but just like the moral liability, I don’t want to be the person they call and say, “Can we hit the enter button on this command or not?” What a crazy dynamic to want to be in. I also don’t have a God complex, so naturally I would not want to be in that position. But if I did have a God complex, then I’d be like, “Yes, you can just call me every time you want to use my software, and ask me if it’s okay to use it.”

Liz Hoffman:
You had Emil Michael on stage, I think last week or two weeks ago. This is the deputy undersecretary who’s in charge of a lot of the Pentagon’s tech use, and he was making the case, essentially that argument, that we cannot, with missiles flying in, be trying to get Dario on his cell to ask him if it’s okay. Do you think that he, A, made that case sort of compellingly in the room, and do you think the government is going to have to make that case to a public that is very wary, and getting warier by the day, of this technology?

David Ulevitch:
I think we’ll find out. It’s hard for me to predict. I think he made a compelling case. I think Dario’s view, which is one that I think is understandable, is that it’s always been the case that technology is often far ahead of legislation. That’s generally an always true statement. I think Dario’s view is that this technology is so advanced, so far beyond other capabilities that have been ahead of legislation, that we actually can’t just rely on legislation catching up, because the window of time between when legislation catches up, and policy catches up, is too wide and the potential consequences are too great. If I was going to make Anthropic’s argument, I think that’s the argument that I would probably make.
The challenge with that is that there has to be another solution. One solution could just be like, I’m not going to sell to the government, which is currently where Anthropic is. You have to recognize that the folks who are making the decisions of how to use this technology, they have all of the intelligence information coming in from the clandestine services. They have all the intelligence information coming from the Department of War, from the special operations groups. They’re the ones ingesting all the satellite imagery that’s highly classified. They’re the ones who have all the foreign information around wiretaps and data of where is Maduro hiding out? What building is he in? They’re using all this information to synthesize a plan that will protect the most American lives and achieve their mission outcomes. Not only is it illegal, but it’s impossible to convey all that information to some third-party extrajudicial person who’s not part of the chain of command. That also doesn’t make any sense either. And so there has to be some other path that I think needs to emerge.
One can be that the technology might be more limited in its capabilities, or there will be some companies that will fine-tune their models to make better decisions in those use cases. We’re not talking about do we or do we not have weapons? What we’re really talking about is, do we have incredibly smart weapons that know the difference between a school bus and a tank, or do we have dumb weapons that don’t know the difference? Do I think AI will ever make mistakes? Of course. Do I think it will make fewer mistakes than humans? I, for sure, think that it will make fewer mistakes.

Liz Hoffman:
Okay. So we’ve covered a lot of the tech. This is a business show, so we want to talk about the business of VC and how it is changing. VC grew up investing in software, at least over the last 25, 30 years, very proudly capital light, infinitely scalable. In large part, sort of consumer-facing business for a lot of the 2010s. Why do you have a right to win in these really hardware-heavy, capital-intensive, strategically complicated businesses?

David Ulevitch:
I will answer your question, but I think there’s a meta answer first, which is that things that were previously held to be sort of unassailable truths are now coming into question. People thought SaaS was the most durable business model of all time, and if you invested in an enterprise SaaS company, it was incredible forever. It would pay out like an annuity.

Liz Hoffman:
Oops.

David Ulevitch:
Then AI came along and said, “Well, wait a minute, the ability to replace your internal IT tools with something that’s perfectly built, the way you want for your own internal system, now is infinitely easier and you don’t need to pay ServiceNow anymore.” These companies that were perceived to be unassailable, I cannot underscore enough how durable people thought these businesses were. Now it’s like, well, these things aren’t even worth really anything. Maybe they’re worth four times earnings, five times earnings, I don’t know, but it’s no longer these 50 times earnings because they’re no longer unassailable.
Then previous conceived beliefs around, “Oh, hardware is so hard, it’s so low margin, it can’t possibly work,” that’s not true. When we start building automation into the factory, when we start taking control of our supply chain, when we start using advanced software to replace things that had to be previously done in software, and I’ll give an example in a second, we actually think that you can get, with high rate of production, you can get high margin outcomes. Look, there’s tailwinds in defense and energy and aerospace. If you look at everything from what SpaceX has done to unlock launch capabilities, to what’s happening in cyberspace, as well as electronic warfare, as well as in munitions, as well as in manufacturing, as well as in energy generation, transmission and storage, the opportunity set just is growing, I would say largely without bound and in a way that I think will be much higher margin than in the past.
I can’t promise that, but that’s what I believe, that’s what we’re betting, that’s what we’re investing against. We’re taking these categories where historically have … No offense to the founders who have worked in these industries for a long time, but we’re now bringing the best Silicon Valley mindset, the best founders into these largely legacy, slow-moving categories and actually applying advanced technology to them, in a way that I think will be much, much more compelling and deliver much greater returns. That’s my answer as to why now. We have this watershed technology moment and the economic demand side is, I would say, just growing without bound.

Liz Hoffman:
That’s something about the way that the hardware is being produced has some DNA that is borrowed from, or grafted from, other things that venture has done in the past.

David Ulevitch:
Yeah. This is one sort of marginal example, but night vision typically has had to use, I think vacuum tubes, really old school vacuum cubes. This is still true today. When you go buy these very expensive exquisite night vision systems, it’s like 10,000, $25,000 and it literally has vacuum tubes in it. I’m not an expert in this, but I’m pretty sure I’m correct about that. New night vision systems, the cameras we have, with all the multiple lenses going on, you can do a long exposure to capture all the light, you can do a quick exposure to get the granularity. Then software basically now can provide almost, or better, in some cases, night vision capabilities without any night vision. You’re just doing it all on software and commodity optics. And so you no longer need this $25,000 optics package when you might be able to use the optics from a $200 phone. So it’s really transformative when you can apply advanced software.
The same thing applies to the factory as well. We have companies like Hadrian that are transforming the way high mix, high rate of production manufacturing is happening. We have companies like Senra that are applying technology to the wire harness business, things that were completely manual labor in the past, but now are being increasingly automated.

Liz Hoffman:
When you say high mix, high rate of production, what we think of as an old assembly line stamping out these little precision parts and then someone’s putting them together.

David Ulevitch:
That’s low mix. Low mix means you only make one thing. High mix means you make lots of different things. If you think about the factory of the future, right now, these factories of the future have robotics and have machines, but increasingly they will become increasingly driven by robotics. You can imagine that overnight a factory could reconfigure itself to make something else the next day and all the robots will just reconfigure their software to redo something else, and do some other process the next day. Today to reconfigure a factory is a massive disruption. You’re changing the way the lines work, you’re changing the way the supply chain and inventory gets managed, but the factory of the future may not be that way. I think the factory of the future just needs incredible ability to carry a lot of weight on the floor and a lot of power.
Right now, factories have power to power the machines, but you actually think about the factory of the future and you’ve got to charge all the robots up every night, or when they’re not working. It’s like electric vehicles, you’re going to need all these charging stations. Your loading dock is not just for bringing in the semi-trucks, the loading dock now is a charging station for all the trucks, and vehicles, and forklifts. It’s going to be very, very exciting, but a lot of things have to change, but when they do, the outcome will be, I think, dramatic production capability.

Rohan Goswami:
We’re going to take a quick break and we’ll be right back with David after this.

Liz Hoffman:
We had Dara, from Uber, on the show two weeks ago, and he was talking about this next trillion dollar opportunity for Uber, which is essentially tucking all the robo taxis in at night, the care and feeding of these big fleets. Super interesting.
Let me just poke at that a little bit. VC has just optimized for totally different things over the last 25 years. The skillset, and the returns that are expected, and the value add that they bring. There are investors out there, mostly public equity investors investing in super CapEx heavy businesses who know how to do this well, and keep executives accountable. Is Silicon Valley convincing itself that it can be good at this because it wants to be because it just ideologically likes this stuff?

David Ulevitch:
No, I don’t think that’s correct.

Liz Hoffman:
Okay.

David Ulevitch:
I would say that there are businesses that we see every single day that we say, “Wow, that’s an awesome business,” but it’s not a venture backable business, because we don’t see the unconstrained upside when we think about the, what if it works, how big can it be? If we were in the private equity business and we could just buy this company and run it, and we could probably run it better than these people, or these other people, or whoever.
Broadly, we think there’s lots of those. There’s an entire category of the sub-subcontractors in the defense industrial base. There’s a ton of people along the way in the manufacturing industrial base where we actually think it’s fine for private equity, and middle market private equity folks to come in and just run these things and really manage to the bottom line. In certain areas, where we think that there is essentially unlimited demand, where there’s transformation moments, and there’s magnetic founders who can attract capital, attract talent, and the market demand side is essentially insatiable, those we think are venture backable. We do think that venture capital is the right place to go there, because we’re really managing to the top line. How fast can you grow?
We know that there are companies we’ve invested in that are very good at spending money. They’re incredible at spending money. The thing that’s also incredible is how fast they can grow their top line. First to get $1 billion program of record, first to get to $10 billion in total contract value, first to ship X numbers of units in a year for a new company. That applies, not just to defense, but also to energy and a bunch of other public safety, a bunch of these other categories. A private equity company could never build Flock Safety, it could never build Anduril because they’re so fixated on the P&L and managing to the bottom line. That’s appropriate for more mature businesses, or businesses that are TAM constrained, but it’s not appropriate.
So there is a place for venture capital. It’s not for every business, but we think it’s increasingly for a larger and larger set of businesses in these categories. Again, we only need to invest in 20 outsized winners. Our loss ratio will be reasonably high, the same way in the other venture capital businesses. Private equity doesn’t like to have a loss ratio. They don’t like to lose money. We’re okay with it, because the winners pay for all the losers and then some.

Liz Hoffman:
It’s interesting, our colleague, Ben Smith, wrote a profile a couple of months ago of Steve Feinberg, the Pentagon official, who’s overseeing a lot of this stuff, who does come from the finance world, but comes from the private equity corner of it. I think it captured some tension between the buy-the-numbers private equity guys and the VC up and to the right dreamers. I think there was an anecdote then a story where he called Anduril like a toy company, essentially. I’m sort of curious what you make of that tension, whether it sounds to you like it is sort of real.

David Ulevitch:
What I would say is that I’ve been in venture now for almost eight years. I’ve been in Silicon Valley since 2004. At some level, the world that we live in has always been people who are willing to press the ‘I believe’ button, and believe that it can become real and people that are like, “I don’t like the ‘I believe’ button, I want the ‘show it to me’ button.” I think it’s okay to have that tension and I think it’s okay to have it in the Department of War. I think it’s acceptable for Deputy Secretary of War Feinberg to say, “I want to see production. I want to see a high rate of production.” What I would push back on is this idea that it’s like primes, versus the startups. The primes are also not in a high rate of production.

Liz Hoffman:
And are getting yelled at aggressively by the Pentagon for that.

Rohan Goswami:
And by the president, for that matter, and focusing on share buybacks instead of manufacturing.

David Ulevitch:
That’s right. That’s right. By the way, it’s not like the share buybacks are the reason they’re not at a high rate of production. They could be in the high rate of production if they had the capability, but they don’t. They don’t have the capability, and they don’t want to, and they have no existential necessity. There’s people in these primes that will say, “I was here before the Trump administration, I’ll be here after the Trump administration. I don’t really care. We ship our 50 missiles a year, leave us alone, whatever. It’s fine.” If I had to bet on which end of the barbell will get to very high rates of production more quickly, is it the legacy incumbent primes who lack the, I would say, both institutional knowledge, or appetite to move very, very fast; or the startups who have just survived an entire period of life of not dying, and are now in the process of moving towards thriving, and are completely fixated, and working all hours of the day, and have the brightest mind magnetically being attracted to them. Are they going to get to high rate of production first?
I would make that bet every single day of the week.

Rohan Goswami:
You talked about missiles just now. You talked about the primes moving slowly. Liz and I covered this a few weeks ago. L3Harris cut this really weird, interesting deal. The DOD is taking a stake in them or taking a stake in their spinoff, and it’s all designed to increase missile production. Do you think they’re capable of actually competing with your guys? They seem pretty front-footed about it. What did you make of that deal?

David Ulevitch:
I think we are in a period of time, especially after the Ukraine war, where we supplied arms, the conflict with Israel and Gaza where we supplied arms, and now the conflict in Iran where we are deploying our own arms. I think we are in a place where we have a massive need to restock the arsenal of freedom, as they call it. I think that everybody needs to get into production capabilities. I think if L3Harris wants to ramp up production, I’m not sure what the government taking a stake in that will do to accelerate it. Maybe there’s somebody who sits in the office just poking them like, “Move faster, move faster.” I don’t know. I also think the startups are highly motivated.
One thing I would say is you have to remember, the startups occupy a tiny amount of the overall defense spending today. If that amount of spending went and doubled from whatever it is, one or 2%, if you take away sustainment programs and personnel programs, the amount of money being spent on new programs with startups is relatively infinitesimally small. If you just doubled or tripled that, it would be a massive demand signal to startups who would actually take advantage of that and respond, in kind, with high rates of production, more capabilities and shipping faster. Giving more money to the primes is not going to help them move faster, because if that were the case, they would’ve been moving faster before.

Liz Hoffman:
You were saying earlier about how you think ventures scales up and your fund is it’s a billion, eight, right?

David Ulevitch:
The new one’s a billion two, but when you add it up with the old one, it’s 1,776 million.

Rohan Goswami:
Very nice. Very nice.

Liz Hoffman:
Because it’s going to be some very finely tuned cash management is what you’re going to do there. It was also part, I think, the last round of a $15 billion fundraising across the firm, which is just, historically speaking for venture, a huge amount of money. History is just full of cautionary tales of asset classes scaling up and ending badly. Understanding that you’re talking about funding really expensive things and funding these startups that are getting bigger and staying longer private, what makes you think that that’s a good idea? At what point are you actually going to bump into the next guys up, which I guess would be private equity, growth equity, right?

David Ulevitch:
Two things I would say is one is the outcomes are just becoming much, much larger. If you look at the proposed SpaceX IPO, and I don’t have any proprietary knowledge about this, I read the headlines and I read the articles, and there’s chatter that it could be worth one and a half trillion dollars when it goes public. I don’t think five, six, seven years ago, anybody thought about how a company could go public being worth one and a half trillion dollars, let alone, by the way, companies like Meta, and Google, and Apple are worth many, many trillions of dollars.
Why can’t there be a defense company that’s supplying advanced technology, at high rates of production, and advanced manufacturing? Can there be one that’s worth a trillion dollars? I think there absolutely can. The outcomes we see are just getting larger and larger. If you look at the energy side of things, whether it’s mining and minerals, whether it’s from a resource extraction standpoint, whether it’s from battery, manufacturing, motors, actuators for robotics revolution, whether it’s for generating energy and things like nuclear power, the market sizes are growing, essentially without bound. Again, it doesn’t even matter if you’re like, “I hate AI data centers. I hate electric vehicles.” You can’t hate everything. The move to electrification is happening and it doesn’t matter which one you’re excited about, the combination of all of them just creates this insatiable demand for energy. The same thing is happening with minerals and mining and the same with manufacturing. We just think the outcomes are larger.
In terms of working with the rest of the capital stack, a bunch of our companies are already doing that. If you have government contracts for a billion dollars to go build 5,000 ships, autonomous surface vessels, you’re going to go work with private capital to finance that, finance the equipment, finance the factory build-out. If you’re building AI data centers and hyperscalers and you have commitments from the hyperscale customers to buy your next 20 years of compute capacity, you’re going to go finance that. You’re going to work with the private credit folks, you’re going to work with the real estate folks, and that’s already happening. You’ve seen JP Morgan commit one and a half trillion dollars to this Security and Resiliency Initiative.

Liz Hoffman:
Sort of their version of American Dynamism Fund.

David Ulevitch:
Except that if you sort of unpack it a little bit, it’s not like they’re committing one and a half trillion off their balance sheet. We’re committing our 1.2 billion, 1,776 million to actually invest in equity in these companies. They’re going to provide banking services, credit facilities, they’re going to be the facilitator of one and a half trillion dollars of economic transactions and business, which is amazing and great. I think they are going to commit a huge amount of money, off their balance sheet, to this. I think that just goes to show that this has to be a partnership, it has to be an ecosystem, but that the actual foundation of it is coming from private capital and particularly venture capital.

Rohan Goswami:
Roelof Botha, who obviously was the steward at Sequoia for some time, is skeptical of your argument that VC scales. To him, he doesn’t think it’s analogous to private equity, or even the public equity markets. He thinks it’s a more bespoke product for higher end investors. What do you say to that?

David Ulevitch:
I think you should ask his partners who made a change inside the firm.

Liz Hoffman:
All right. We’ve got our next booking assignment.
A16Z, you have a big public safety portfolio. You are on the board of Flock Safety, which was, I think, at the sort of the heart of that Amazon Ring commercial that was slightly dystopian and comically backfired a little bit, but a bunch of local communities in the last couple of weeks have backtracked from some plans to install public cameras. There’s a real anxiety, I think about living in a Panopticon world, even if it maybe helps you find your dog. Are you feeling the backlash to that, or is this kind of a one-way street that you think we’re on to widespread surveillance in the name of public safety? Where do you draw the line there?

David Ulevitch:
I think it’s an incredibly important discussion/question. I’ve thought about privacy in my entire life. When I was 16, I would go to DEFCON, the hacker conventions. This is all I was fired up about. I used to host all these websites that were privacy focused. I got sued by Diebold Corporation, which made voting machines.

Liz Hoffman:
Wait, really? Hold on. Hold on. We’re going to need to know more about that. What did you do?

David Ulevitch:
I mean, this is really a tangent. I was hosting websites as a project called the California Community Co-location Project, and we hosted a bunch of websites. We hosted people like Kevin Sites who were journalists exposing things happening in the Iraq War. I hosted RiseUp.net. I used to host the DNS for WikiLeaks. I hosted lots of things. I just sort of believe in freedom of speech very strongly. One of the things we hosted published documents that showed that the Diebold voting machines were insecure, or had some insecurities. Diebold, they came out publicly in the press and said, “Well, these are fake documents. They’re not real.” Then they sued me for copyright infringement. It’s hard to know, if something’s not real, how could it be copyright infringement? It turns out that was a bad strategy for them and they ended up settling out of court.

Liz Hoffman:
Amazing.

David Ulevitch:
It actually ultimately ended up being very good. Yeah.

Liz Hoffman:
You undersold it. That was a good tangent.

David Ulevitch:
Anyway, I do think privacy is really important. What I think the issue is, is we all walk around with phones in our pocket that are constantly tracking our location, that’s being sold to data brokers. Our privacy is being violated in ways that people cannot even possibly, I think, fully imagine and understand, and with no ROI for the individual citizen. Now, when it comes to license plate readers, this actually solves crimes and gets criminals off streets. This is actually protecting people and making them safer at what I would say is very, very, very little cost to the citizen. Again, I’m concerned about the Panopticon. It’s not coming from Flock Safety, it’s coming from the data broker infrastructure that underpins the cellular network and the social media networks and those things. That’s actually where it’s coming from.
I’m not trying to do a, “Hey, look over there,” kind of thing. I’m really not trying to do that. But people, because they can’t conceptualize and point to a thing, when it comes to the data brokers and how that privacy is being violated, they point to something like Flock Safety. Now, what I will say is that the way Flock Safety works is, again, it’s technology and they sell to jurisdictions that have their own rules and laws. Some jurisdictions have no data retention policy. Some say, “We’ll keep data and photos for seven days.” They audit and log all the searches, so they know if a law enforcement officer is doing a search to solve a crime or searching and looking up an ex-girlfriend. They know that, and then those people get busted and they get fired, they go to jail, whatever.
There’s a tremendous amount of oversight, and audit and legal compliance around these things. I really do think that in this case, it’s much ado, not about nothing, but much ado about the wrong thing. They get poked at, because they’re tangible. You can point to a Flock safety camera and say, “Oh, there it is.” The reality is anything but that.
In fact, there was an example a couple of days ago in Washington state where a community had … I think there was a shooting, or a murder, that took place and they had recently removed their Flock Safety cameras, and this is in Richland, Washington. Then the cameras in Olympia, Washington actually caught the suspect. So the neighboring jurisdiction, which kept their cameras, was actually able to find the suspect and get this person pulled over, and arrested, and helped solve a crime in the neighboring jurisdiction.
I think that we’re in a moment in time where people care about this. They should care about it. We should talk about it. I think in public safety, the benefits far, far outweigh the concerns. There’s so much oversight. It’s really the wrong thing to focus on when we talk about the surveillance date, or the Panopticon, or these other issues.

Liz Hoffman:
It’s interesting though, you’re talking about there’s oversight and we’ll cut to the Anthropic of it all. That is very powerful technology.

David Ulevitch:
Totally.[inaudible 00:39:08].

Liz Hoffman:
The question is how much did the company control the use of it? There was some reporting a week or two ago that, I think it was OpenAI had been reviewing the chats of one of the mass shooters.

David Ulevitch:
The humans decided not to call the police.

Liz Hoffman:
Correct. It just feels like we’re just going to have this really big fight and discussion. I don’t know, I’m not saying that the line needs to be exactly the same for Anthropic and the Pentagon as it is for Richland, Washington and Flock. Where do you see that all shaking out?

David Ulevitch:
I actually think it is a very similar discussion. As you see, some jurisdictions will say we’re okay with more crime happening here without having license plate readers. Some jurisdictions will say, “Wait, we don’t want crime here and we are going to trade-off having license plate readers.” They’re going to make that decision and you’re going to see how voters react. I think that a lot of people will choose to live in a community that says, “We would rather have occasional license plate reading, with seven-day retention, as our local jurisdiction allows, and have less crime than saying no license plate reading and more crime.”

Rohan Goswami:
Talk to us a little bit about space. I mean, obviously you guys invest in there, but how do you think about it generally and how it fits into the other industries that you’re playing in?

David Ulevitch:
The only thing I would say about space is that domains that we previously thought were discreet and separate; energy, defense, space are so much more intermingled and related than I really appreciated. They depend on each other tremendously. So without abundant energy, you really do not have national security. Without reliable space-based communications and space domain awareness and space-based capabilities, you do not have strong national security. These things all play a part together. Without strong energy, you can’t have a manufacturing and defense industrial base. These things are very, very interrelated. The Venn diagram is almost entirely overlapping, from a resilience and national security standpoint.
We have to have all these key industries, I think, supported and they’re all very important. We’re going to go through a major technological transformation, but it’s going to unlock just incredible opportunity. People fixate and focus on AI will cause these jobs to be lost, or data centers are going to consume a lot of power, but we’re going to have more power. We’re going to have entirely new economies and things created from AI and from manufacturing. To me, we’re entering just a very exciting world where things are going to all be rapidly accelerated by technology.

Liz Hoffman:
I think we skipped over when we were talking about whether venture can, and should be, in the business of doing really expensive, heavy, dangerous things. You are on the board of a nuclear company. A lot of talk about nuclear now, particularly as there’s just massive energy shock coming in from Iran, but also just a huge giant sucking sound just coming from AI off the grid. Where do you see, in five or 10 years, is nuclear a real meaningful part of the new build in the US?

David Ulevitch:
I think that is the case. Just on the Iran energy shock, there was an article in the Wall Street Journal from the editorial board about thanking Paul Ryan for allowing the US to export crude, which has actually softened the blow of rising oil prices and stabilized prices much more than they would be otherwise. That said, there is this insatiable thirst for energy. I talk about this all the time. It doesn’t matter if it comes from solar, from fossil fuels, or from nuclear, but the thing about nuclear that I think people don’t realize, because it was maligned for so long, was that, one, the number of people who have died from nuclear accidents in this country since the beginning of nuclear capabilities I think it’s approximately zero. By the way, the people that die from coal mining, just the long-term consequences of burning fossil fuels, has got to be in the millions of people.
So we’re like, zero people have died versus millions of people have died. I think that people no longer see it as a threat. Then the question is, can we actually generate nuclear power economically? The answer to that is largely by … I think actually AI will play a role in that the regulatory hurdles for nuclear are very, very onerous. There’s a tremendous amount of paperwork and permit process, some of which is justifiable, but AI can actually rapidly, rapidly decrease the time to apply for regulatory permits, for nuclear fuel permits, for transport permits, for manufacturing permits. Then on the government side, the AI can actually help them process the applications much more quickly. That’s actually happening today. We are seeing the supply chain for nuclear fuel be reshored domestically. We are seeing the manufacturing of nuclear facilities be shored up domestically. We now have a factory in Oak Ridge, Tennessee for radiant nuclear that’s making nuclear power plants.
These are transportable and they fit in a shipping container kind of thing. Then we see large-scale nuclear being done, and all of them are very, very good. We should be doing all of them. There will never be a point in the United States where we have too much power generation. That will never happen. It’s like compute, it’s like storage. As we generate more storage, we just find more data to put in it. As we generate more compute, we find more use cases for it. The same thing is true for power. If we generate more power, we will find more use cases for it. We will electrify more vehicles. We will power up more data centers. We’ll clean more water. We will clean more air. Again, we have an unlimited ability to use any amount of power that we generate.

Liz Hoffman:
By the way, I grew up two miles from Three Mile Island.

David Ulevitch:
You are alive and you appear healthy.

Liz Hoffman:
10 fingers holding. If anyone’s going to be particularly twitchy about this, it might be me. I’ve just always thought that, particularly in parts of Europe, even the US, just a real self-inflicted wound. I’ve never totally understood the reflexive panic around nuclear. It just seems like a pretty great technology.

David Ulevitch:
I think the Sierra Club is largely responsible for some of that.

Liz Hoffman:
They would like there to be fewer emissions, this is the cleanest technology we’ve got.

David Ulevitch:
Totally.

Liz Hoffman:
It’s interesting.

David Ulevitch:
Again, look at a country like France. They have never had a nuclear disaster. They’re very, very modern when it comes to nuclear technologies. We are too. By the way, we put nuclear reactors in boats, and in our aircraft carriers, and some of our submarines. We have nuclear submarines. It’s never been a problem. Nobody’s died. In fact, most of the fuel that people are using for modern reactors, you could hold in your hand, you’d be totally fine. No issue.

Liz Hoffman:
You can hold it in your hand, David. I think I will not do that.

David Ulevitch:
Yeah. That’s fine. I wouldn’t recommend eating it, but I think you’d be fine to hold it.

Liz Hoffman:
I’m sure they’re shaving it on truffle fries in Midtown.
You were saying earlier that there’s grumpy people at the prime saying, “We were here before Trump. We’ll be here after Trump.” Where will you be after Trump? Scott Alexander, this influential tech blogger, was kind of giving you guys a hard time and saying that as soon as the Democrats are back in power, Silicon Valley writ large will go back to thinking government is evil. What do you make of that?

David Ulevitch:
I think our firm is very bipartisan. I think that people can say, “Oh, this is only happening because of the Trump administration.” I think what the Trump administration has done is show that it’s okay to make change, and it’s okay to push for change, and not get too mired. Look, there will be some downside consequences. I think it will be wildly positive, sort of writ large, but there will be things that will not go perfectly and we’ll have to course correct. That’s always been the case, and we might as well at least do it quickly and get their fastest.
I hope that that carries on to the next administration, no matter who it is, that this idea that we have to change procurement, we have to change the way that we apply advanced technology, they have to change the way we do contracting, we have to have the best founders building for these missions. That should be bipartisan. That is the thing that America does the best. We are the best at it. It doesn’t matter whether it’s operation warp speed or defense production. American sort of creativity, entrepreneurship, ingenuity and resilience has always been the thing that we’re best at and we should lean into that at all times.

Liz Hoffman:
That feels like a great place to end it. Honestly, David, this was a lot of fun.

Rohan Goswami:
This was great.

Liz Hoffman:
Thanks for doing it.

David Ulevitch:
Thanks. My pleasure.

Rohan Goswami:
All right, Liz. We’ve talked about this before. You’re a little skeptical, generally, that venture has any right to play in, like you said to David, this CapEx really intensive space. Did he change your mind?

Liz Hoffman:
Yeah, a little bit. Yeah. I think so. I mean, we’re thinking about these capital-intensive industries. I think maybe I had kind of an outdated view of them, particularly on the factory side. He was talking a little about precision components and this sort of, we call them, mixed use high volume. What we think of as an assembly line, I would say they have absolutely no right to play there, nor would they want to, but that the sort of manufacturing base is way more tech enabled and software enabled than most people realize, and that may give them a right to play there.
I think where I remain skeptical, and maybe I’m scarred here, I spent a couple of years covering the SoftBank Vision Fund, which was orders of magnitude larger than any VC fund that had ever been raised, and did not go great.

Rohan Goswami:
Yeah. They sucked at deploying capital.

Liz Hoffman:
Well, right, because I think, and maybe this is where I would steal man his argument, they were deploying the 2026 amount of VC money on 2017, 18, 19 vintage products.

Rohan Goswami:
Yeah. His point is that if you think about what the exits look like, a trillion dollar exit, a 10X, a 500X exit, it’s reality now. You’re seeing that test sit in the public markets.

Liz Hoffman:
Yeah. Throwing $10 billion at WeWork, which is, I was saying that old generation of Uber, but for this other thing and these marketplaces, super CapEx light, that is just a recipe for them to buy insane things and throw insane parties. Whereas investing it in companies that have real use for it, is different.
I think then the question comes down to, do they have the right operators at these companies? I remember talking, a couple of months ago, to the CEO of one of the big oil companies and he’s like, “Yeah. These guys don’t know how to do CapEx. We do CapEx for a living.”

Rohan Goswami:
How to spend their money right.

Liz Hoffman:
Correct. He was like, “They’re absolutely going to waste 10% of this money.” Maybe that’s a trade-off that, as a country, we should be okay with.

Rohan Goswami:
Right. If they can just overrun by 10%, okay, they’re not an oil field, they’re not a telecom. There are no constraints on how they spend capital. The government foots the bill no matter what. It is kind of the perfect business for venture in that the upside is unlimited.

Liz Hoffman:
Totally. It’s interesting, right? You’re trading an infinite number of customers for kind of one customer.

Rohan Goswami:
Yeah.

Liz Hoffman:
You wonder what will happen. There’s been a lot of talk is there are going to be a, what is a sixth prime or seventh prime, Palantir has cast itself as that. You wonder if Anduril breaks into this club, are they going to be quite as dismissive of the process that protects the primes? Yeah, exactly. That’s just a sort of a co-op problem that they’ll solve when they get there.

Rohan Goswami:
I was reminded when he was talking about the factories of the future, these mixed use, high volume places. I was reminded of our conversation with Dara from a few weeks ago. The question for me there was like, where are the humans in all this? If everything is autonomous, if all the drones are AI swarms, where are the humans in all this? Are they even programming anymore? What do you think of that?

Liz Hoffman:
I don’t think that Silicon Valley spends a lot of time thinking about that. Actually, I’m down here for BlackRock at a big infrastructure summit yesterday. I was interviewing the CEO Larry Fink about this and they’re making a real pivot into blue collar jobs, electricians and welders. We do not have enough of them. We didn’t have enough of them as a country before we started this giant AI data center boom. These are good high paying jobs, but you’re starting to see … Also, Jensen Huang, CEO of Nvidia, put out an essay the other day about AI. There’s a couple of paragraphs in there that I think we’ll revisit them. It was an acknowledgement that this is scary for labor and we’re still trying to figure it out.

Rohan Goswami:
I didn’t see this.

Liz Hoffman:
It’s super interesting. It’s worth reading. By way of saying, I think historically, Silicon Valley has not spent a lot of time caring, but I think they’re going to find themselves on the front lines because the jobs that are really starting to go are obviously the tech jobs and the knowledge economy jobs. I think they’re going to have to find a messaging strategy because otherwise this ends with perhaps metaphorical, but heads-on pikes, they’re starting to think more about it. I think that that is smart and overdue.

Rohan Goswami:
I mean, look, certainly my generation was told, many of them, “Go get a computer science degree, work in the valley, make a ton of money. Join a startup, found a startup.” I obviously opted for a slightly different path than that, but there was this social compact that this is the way to success.
I worry about the robots, Liz. I am very scared about the robots, because there’s no mode against those guys even for the blue collar workers because eventually it’ll just be robots servicing robots in space.

Liz Hoffman:
Yeah. Totally. We didn’t ask him squarely, but it’s pretty clear that he thinks a lot of the future of war is unmanned. On the one hand, that’s great, but then the question is, well, if a human cost of war goes down, we’re going to have a lot more war. That is how markets like that function.

Rohan Goswami:
The check has always been the electorate. People don’t want to send their kids to war. They don’t want to see their friends and family die.

Liz Hoffman:
Yeah. Just as an economic matter when the cost of something goes down, you get more of that thing. I don’t know. He was super interesting. I think he was more honest actually, than I would’ve thought, about the politics of this. We’ve seen all these companies rejigger themselves in some sort of MAGA way. Sometimes it’s a little gratuitous and obsequious, but sometimes they’re just doing things, that in a different era, would not have felt coded to one party or another. David obviously sits at Andreessen Horowitz and there’s no doubt about particularly how Mark Andreessen feels about politics, and Biden, and Trump. I don’t know, it was sort of hardened by what he said that he thinks that the stuff will outlast this administration, or perhaps retreat from its incredible partisan coding. I’m hopeful, I’m not maybe as optimistic as he is.

Rohan Goswami:
Well, that’s it for us this week. Thanks for listening to Compound Interest from Semafor Business. Our show is produced by Josh Billinson.

Liz Hoffman:
With special thanks to Anna Pizzino, Katherine Bilgore, Matthew Alexander, Claire Einstein, Rachel Oppenheim, Tori Kuhr, Vilanna Wang, Garett Wiley, and Daniel Hoeft.

Rohan Goswami:
Our engineer is Bob Mallory. Our theme music is by Steve Bone.

Liz Hoffman:
If you like Compound Interest, please follow us wherever you get your podcasts and leave a review. If you want more Semafor Business, you can sign up to get our email briefings in your inbox.