Marc Andreessen——技术霸权之战——[Invest Like the Best, EP.410]

My guest today is Marc Andreessen. Marc is the co-founder of Andreessen Horowitz and one of Silicon Valley's most influential figures. He combines deep technical knowledge from his engineering background with broad historical understanding and strategic thinking about societal patterns. He last joined me on Invest Like the Best in two thousand twenty-one, and the playing field looks a lot different than today. I've made sure to try to listen to all of Marc's many recent great podcast appearances to make sure that we covered very different ground than Marc has covered elsewhere. He goes deep on the seismic shifts, reshaping technology and geopolitics. We discuss DeepSeek's open source AI and what it means for technological rivalry between America and China, his perspective on the evolution of power structures, and the transformation of the venture capital industry as a whole. Please enjoy our conversation. And if you haven't yet heard, last week at Colossus, we formally launched a subscription to our new property, Colossus Review. It's a quarterly print, digital, and audio publication that profiles the investors, founders, and companies that we respect most in more detail than we've ever gone before. To learn more, go to joincolossus.com/subscribe or click the link in the show notes. Marc, I think we have to start at the white hot center. Can you just riff on your reaction to DeepSeek's R1?

1. DeepSeek Makes AI Ubiquitous

There's a ton of dimensions to it. It's a really big deal. The US is still, by far and away, the science and technology leader in AI, and so most of the ideas in DeepSeek are derived from work that's been done in America or in Europe over the last 20 years—or, actually, amazingly, 80 years. The original work on neural networks was done all the way back in the 1940s at American and European research universities. From an intellectual development standpoint, the US is still way ahead.

But what DeepSeek is is a really, really good implementation of those ideas. Then they did this marvelous thing, which is that they gave it to the world in the form of open source. It's actually fairly amazing that this has happened. There's been kind of an inversion, because you have these American companies with names like OpenAI that are basically completely closed. Part of Elon's lawsuit against OpenAI is that he's demanding they change their name from OpenAI to ClosedAI.

The original thesis of OpenAI was that everything was going to be open source. Everything's been closed down, and these other big AI labs like Anthropic are also completely closed. In fact, they've even stopped publishing research. They've really taken everything proprietary. The DeepSeek guys, for their own reasons, are delivering on the promise of actual open AI, actual open source.

They've published both their LLM, which is called V3, and their reasoner, which is called R1—the 2 parts of their system. They published the code and technical papers that document in detail how they built it and basically serve as a roadmap for anybody else who wants to do the same kind of work.

It's out, and there's a fake narrative out there that basically says, if you use DeepSeek, you're giving all your data to the Chinese. That is true if you use DeepSeek on the DeepSeek website, if you use the service the way they run it. But you can download the code and run it yourself.

I'll just give you an example of that. There's an app called Perplexity that's now quite popular, an American company. You can use DeepSeek R1 on Perplexity, completely hosted in the United States. Both Microsoft and Amazon now have cloud versions of DeepSeek running where you can run it in their clouds—obviously, both American companies with American data centers.

This is really critical. You can download the system now and actually run it. You can't run the full version yet on your laptop, but you can run it on $6,000 worth of hardware in your house or in your business. It has comparable capabilities to the leading-edge systems from OpenAI and Anthropic. Those companies spent many, many multiples of that amount to build their systems. Now you can run it on $6,000, and you have total control.

If you're running it yourself, you have total control and total transparency into what it's doing. You can modify it and do all kinds of things with it. Then it has this characteristic that works really well called distillation, where you can take the big model that takes $6,000 worth of hardware and create smaller versions of it.

People online have already created smaller versions of it and built them so that you can run them on your MacBook or your iPhone. Those versions are not quite as smart as the full version, but they are quite smart, and you can create custom-tailored, distilled versions that are smart at specific things.

This is a very big advance. It takes both LLM reasoning, which is creative reasoning, and R1, which is actual reasoning for things like math, code, and science, and makes them something that was super-esoteric, ultra-expensive, and proprietary 6 months ago into something that's universally available to everybody for free forever.

Every major tech company, internet company, and startup is working on this. We have tons of startups—dozens or hundreds of startups this week—who are either working on replicating DeepSeek, incorporating it into their products, or figuring out the techniques that they used and applying them to make existing AI systems better.

The team at Meta—Mark Zuckerberg talked recently about this—is ripping it apart, taking the ideas totally legally because it's open source, and making sure that the next version of Llama will be at least as good at reasoning, or better.

This really propels the world forward. The 2 big things that we can derive from that are, first, that AI is going to be ubiquitous. There are all these AI-risk people, safety people, regulators, bureaucrats, the government, the EU, and the Brits with their safety thing. There are all these people who want to lock down and control AI, and this basically guarantees that none of that's going to happen, which I think is great.

It's very much in the tradition of internet freedom and free speech, so I think that's fantastic. This was a 30x cost reduction for the same capability in 1 step.

Maybe the final thing to note is that this shows reasoning is going to work. Reasoning is going to work for basically any domain of human activity in which you can generate answers that can be checked after the fact to determine whether they're correct by a technical expert.

We're going to have AI that's able to do human- and superhuman-level reasoning for real, and that's going to work in very important domains: coding, math, physics, chemistry, biology, economics, finance, law, and medicine. This basically guarantees that every human being on the planet, in 5 years, is going to have a superhuman AI lawyer and AI doctor at their beck and call at all times, just as a standard thing on their phone. It's going to make the world a much better, healthier, and more wonderful place.

Obviously, this is the least static story of all time, and within 2 months, this model will be stale. Other models will come out, with lots of innovation happening at every level of the stack. But just zooming in on today and this new paradigm that we've entered, if you were writing a winners-and-losers column of all the various stakeholders—whether that's new app developers, incumbent software developers, infrastructure providers like NVIDIA, closed-source versus open-source model companies, and so on—who do you view as the major winners and losers post-R1 today?

2. The Winners And Losers

To your point, if you're taking a snapshot in time today, we'll start with that. If you're looking at a zero-sum game, with winners and losers just at a point in time, I would start with the winners.

The winners are all the users: all the consumers, every individual, and every business that uses AI. We have all these startups doing AI legal and AI lawyer work, and last week they were spending 30 times as much on AI as they're spending this week.

For example, take a company building an AI lawyer. If the cost of your key input drops by 30X, it's like driving a car and the cost of gasoline dropping 30X. All of a sudden, you can drive 30 times as far on the same dollar, or you can use that additional spending power to buy more things. All of these companies are going to be either dramatically expanding the capability of what they can do with AI in all these domains, or they're going to be able to offer their services cheaply or for free instead. And so, for all the users and for the world, it's great.

On a fixed-size-pie basis, the losers are the proprietary model companies: OpenAI, Anthropic, and so forth. You'll notice both OpenAI and Anthropic put out pretty strong but clearly provoked messages in the last week explaining why this wasn't terminal for them. There's an old adage in business and in politics: “When you're explaining, you're losing.” They definitely noticed this.

The other is NVIDIA. There's been a lot of commentary on this, but NVIDIA makes the standard AI chips that people use. There are some other choices, but NVIDIA is the one that most people use. The margins on their chips are as high as 90%, and the company's stock price reflects that: It's the most valuable company in the world.

One of the things the DeepSeek guys did and documented in their paper is figure out a way to use cheaper chips. They're actually still using NVIDIA chips, but they're using cheaper chips and using them much more efficiently. Part of the 30X cost reduction is that you just need a lot fewer chips. And then, by the way, China's building its own supply chain of chips, and these guys are also starting to use the Chinese-derived chips, which is, of course, an even more fundamental threat to NVIDIA.

So that's the snapshot at a point in time. But the implication of your question is that there's another way to look at it, which is over time. Over time, what you want to look at is the elasticity effects. Satya Nadella used this phrase called the Jevons paradox.

Think of this like gasoline. If you drop the cost of gasoline dramatically, then all of a sudden people are going to drive a lot more. This comes up in traffic planning a lot. You'll have a city like Austin, and it'll be gridlocked, and somebody will have the bright idea of building a new freeway alongside the current freeway. Within 2 years, the new freeway will have also filled up with traffic, and maybe it'll even be harder to get from place to place.

The reason is because a reduced price of key inputs can induce demand. If AI is all of a sudden 30X cheaper, people might use it 30X as much. Or, by the way, they might use it 100X as much or 1,000X as much. This might mean that AI gets built into many more things than it was going to get built into before, and much more quickly. The economic term there is elasticity. Falling prices equal an explosion of demand.

I think there's a very plausible scenario here that, on the other side of this, usage explodes, DeepSeek does really well, and, by the way, so do OpenAI and Anthropic. By the way, so does NVIDIA, and so do the Chinese chip makers. Then you have a rising-tide effect where the entire industry explodes in size.

We're really only at the beginning of people figuring out how to use these technologies. Reasoning just started working in the last 4 months. OpenAI released its o1 reasoner not that long ago, a few months ago, and then—

Yeah, it's incredible.

This is like fire being brought down from the mountain and being given to humanity, but most of humanity is not yet using fire. They're going to. I tend to be a believer that it's much more of a—at least for a while here—it's going to be much more of a positive-sum thing, where there's just going to be so much explosive growth.

Part of this, quite frankly, is the old idea of creative destruction. It's like, okay, if you're OpenAI or whatever, whatever you were doing last week is no longer good enough. But, by the way, that's the way of the world. You have to get better. These things are all races. You have to evolve. This is also a very big motivating catalyst for a lot of the current companies to really sharpen their game and get more aggressive.

One of the things that's so interesting is how immediately the mind switches. I think you called it the Sputnik moment. One axis of comparison is Americans versus the Chinese; it's a new race, not so much Anthropic versus OpenAI versus Llama or whatever. How do you think about that axis of competition and comparison, and what will it result in? What is the Sputnik moment? Maybe explain that idea and riff on that new source of competition a little bit as well.

3. The New AI Cold War

The Sputnik moment metaphor comes from the first Cold War, the 20th-century Cold War between the U.S. and the USSR, and a lot of your listeners are too young to remember that. I grew up at least in the tail end of that. I remember it quite clearly as a kid. It was still going hot for the first 18 years of my life. I remember in high school in the 1980s literally thinking there was a good chance that we were all going to die in a nuclear holocaust. The whole nuclear weapons thing—tensions ran super high during the 1980s. There's a great TV show called The Americans that recreates what it was like at that time. It was tense. Things were not good.

Basically, the USSR peacefully collapsed in one of the more amazing twists and turns of history in 1989. The Cold War never became hot, but it was pretty intense. There was this very aggressive rivalry between the U.S. and the USSR through that period. That rivalry was military in many ways, including proxy wars all over the world, but it was also economic.

Many, many Western experts insisted for decades that communism was better than capitalism. American economics textbooks as late as the 1980s said that, inevitably, the Soviet economy would win because state planning and central planning were clearly better than the messy process of capitalism. There was this massive economic battle playing out.

There was also an ideological battle playing out, which was basically freedom and democracy—critique the West however you want, but Western hallmarks of freedom and democracy and free speech and so forth—against the Soviet totalitarian system. The thing about that, though, was that economies were just simpler in those days, and technology was simpler.

The main outputs of the USSR in that era were basically wheat and steel, which is to say basic agricultural exports and basic raw-material exports: oil and relatively simple commodities. By the way, part of the reason the USSR lost is because the U.S. raced way ahead on technology, particularly in computing.

Starting in the 1970s, the computer took off. The Soviet system—the Soviet communist, centrally planned system—was unable to build its own computer industry. That meant that it was ultimately doomed basically the minute the microprocessor was invented, and it just took about 20 years for that to play out. Anyway, there was a technological fight that also happened.

The dynamic between the U.S. and China is similar in many ways, but not completely the same. Niall Ferguson and others have talked about the U.S. versus China as Cold War 2.0. There's competition along each of those dimensions. There are very different worldviews, very different political systems, and very different economic systems.

Both the U.S. and China want to proliferate their philosophy, their method of government, their technology stacks, and their industries around the world. You may recall there was this whole thing that played out over the last decade with a Chinese company, Huawei, that was going to all these other countries and basically selling them telecom equipment to build broadband and wireless networks using a Chinese stack.

The Chinese tech stack for building a telecom network was something like a third the price of the Western stack. There has been this pitch battle playing out globally where both the U.S. and China are trying to get the world on their tech stack.

Why would China want to do that? One reason is because Huawei is a big export industry, exporting routers, switches, and the things that they make. But the other is simply control. If a country bases its wireless network on Huawei, then the Chinese have the ability to backdoor the network anytime they want. They can listen in on any telephone conversation happening in whatever country implements that.

They have a massive geopolitical, national-security, and intelligence reason to proliferate this. The Chinese state puts a lot of weight behind its national champions to go out globally and do this. We saw that playing out over the course of the last decade.

AI is the next version of that. AI is the next thing that America, especially under this new administration, is focused on. The last administration was conflicted on this, which we can talk about. But under this new administration, the Trump administration, the U.S. has a very clear national-security interest and priority goal. The government has a very clear goal to make sure that U.S. technology is the global standard for all the reasons that we discussed.

The Chinese also have that goal, and they have these programs with names like Digital Belt and Road and so forth, where they're going out and doing this. AI is the next turn on that. Every country in the world is going to be buying and adapting AI into its healthcare system, education system, transportation system, telecom system, computing fabric, and energy system.

AI is going to get infused into all these systems that run countries. There's going to be this big fight over whether that's U.S.-derived AI or Chinese-derived AI.

The assumption, of course, up until basically last week, was that China ran a closed political and economic system with a lot of state control and top-down control. The US ran an open, democratic, bottoms-up free-market system. So the assumption up until last week would have been that, of course, we Americans were going to have the more open and freedom-oriented approach to AI, and of course the Chinese were going to come in with a much more closed, controlled, censored version.

Like we said earlier, at least right now, the funny thing—not funny “ha-ha,” but the funny, ironic thing—is that that has flipped. Sitting here today, China took the lead last week in having the best open-source system. They’ve kind of done a judo move. They’re kind of using our strengths against us.

It’s net good for the world because everybody gets to download and use their stuff, and it’s fantastic for all the reasons I described. And so this is far from a hostile act on their standpoint. It’s much more of a generous act. But it’s the kind of generous act that you would expect the West to do, not communist China to do.

And then in the US, we have to decide how we’re going to respond to that. AI policy in the West has been very screwed up because we have not been clear-eyed on this. The way I describe it is, over the last several years, I got very involved in the AI policy debates in the US and in the UK, and then from a distance in the EU.

What I would describe is very fuzzy, confused thinking, where you had a lot of people in power basically thinking, “Oh yes, if it’s a competition with China, we have to win. But if it’s not a competition with China, then we can adopt a Chinese model, and we can lock it down in our own societies. We can use AI as an instrument of control and censorship in our own society.”

So as long as China is not a challenge, essentially the US, the UK, and the EU are going to become more Chinese. They’re going to have more centralized control, more restrictions, more lockdowns, and more censorship. The Chinese are doing us a favor in an ironic way, which is they’re kicking us in the butt and basically saying—I’m going to impute something to them that they haven’t said—but basically, the message is, “You Westerners are betraying your core principles. You’re letting us do this the way that it should be done, as opposed to you guys doing it.”

“If you in the West don’t adapt and actually return to your core principles of freedom, democracy, and openness, China’s not going to just let the world coast along. China’s going to go ahead and seize that mantle.” I think it would be ruinously destructive for the West if that’s what happens. I think we need policies, companies, and products that respond to that in the right way.

It seems like such a fascinating and complicated scenario where they’ve released an open-source model, while the closed—say, Anthropic and OpenAI—companies in the US, which have been the pioneers in this space, are saying, “Look, this model wouldn’t exist if it wasn’t for us. It is being built on top of our work.”

I’m curious about your reactions to IP and the ability for them to do that. I have both a theoretical or ideological curiosity here and a practical one. If you’re Anthropic, let’s say, how are you approaching this if you’re running that business?

What are your ideas, ideological and practical, about a Chinese company using models that were developed here in the US with lots of capital to make the thing that then leads to this abundance for the world? It’s such a weird thing to wrap one’s mind around, and I’d just love your reaction through those 2 lenses.

4. The Open Source IP Battle

There are some real issues here. There is an irony to the line of argument, and you do hear that line of argument. The irony, of course, is that OpenAI did not invent the transformer.

The core algorithm of a large language model is something called the transformer. It was not invented at OpenAI. It was invented at Google. Google invented it and published a paper on it and then, by the way, didn’t pursue it.

They continued to do research on it, but they didn’t productize it. They didn’t productize it because of, quote-unquote, safety. They were like, “Ah, this could be unsafe, and so let’s not do it.” So they let it sit on the shelf for 5 years, and then the OpenAI guys figured it out, picked it up, and ran with it.

Anthropic is a spinoff of OpenAI. Anthropic also did not invent the transformer. Both of those companies, and every other American lab working on large language models, and every other open-source project, are built on something that none of them actually created and developed themselves.

By the way, Google invented the transformer in 2017, but the transformer itself was a derivation of the idea of neural networks. Neural networks are an idea that goes back literally to 1943. So 82 years ago is actually when the original paper on neural networks came out, and the transformer built on 70 years of research and development, much of it funded by the federal government and by European governments at research universities for a very long time.

This is one of those things where there’s an incredibly long lineage of intellectual thought and development that’s gone into it. The vast majority of the ideas that go into all these systems were not developed by the companies that are currently building the systems.

Nobody sitting here, including any of our companies, has any special moral claim that somehow we built that de novo and should have complete control over it. It’s just not true. I would describe arguments like that as frustration in the moment.

By the way, they’re also kind of moot-point arguments, because China went ahead and did it. It’s out. It’s done.

There is an argument around copyright. If you talk to experts in the space, a lot of people have been trying to understand why DeepSeek is as good as it is. One of the theories—and this is an unproven theory—is that China probably trained on data that the US companies don’t train on.

In particular, one of the really surprising things about DeepSeek is that DeepSeek is really, really good at creative writing. DeepSeek is probably the best creative-writing AI in the world right now in English, and this is fairly weird because China’s a different language.

There are some very good Chinese novelists in English. Generally speaking, you’d expect the best creative writing to be coming out of the West in English, and DeepSeek is today probably the best. It’s shockingly good.

One of the theories is that DeepSeek trained on, for example, these websites with names like LibGen, which are basically giant internet repositories of pirated books. I myself would never use LibGen, but I have a friend who uses it all the time.

It’s like a superset of the Kindle Store. It’s got every digital book, and it’s up there as a PDF, and you can download it for free. It’s like The Pirate Bay for watching movies or something.

The US labs might not feel like they can just download all the books from LibGen and train on them, but maybe the Chinese labs feel like they can, so there’s potentially a differential advantage there.

That said, there is this looming copyright fight. People need to be careful in how they think about this, because there are certain publishing companies that would basically like to prevent generative AI companies like OpenAI, Anthropic, and DeepSeek from being able to use their content.

There’s one argument that says that material is copyrighted and can’t just be used willy-nilly. There is another argument that basically says, when an AI trains on a book, it’s not copying the book; it’s reading the book.

It’s the AI equivalent of reading a book, and you and I are allowed to read a book. By the way, we’re allowed to check a book out from the library. We’re allowed to pick a book up off the street. We’re allowed to read a friend’s copy of a book.

Those are all legal. We’re allowed to read the book. We’re allowed to learn from the book, and then we’re allowed to go about our daily lives and talk about the ideas that we learned in the book.

There’s another argument that says training an AI is much more analogous to a human being reading a book, as opposed to stealing it. Then there’s just this practical reality: if China doesn’t care about US copyright, which they traditionally don’t, and their AIs can get trained on all the books, and if the American companies are prohibited ultimately by law from being able to train on the books, then the US may just lose on AI.

From a practical standpoint, that may be a death blow, where it’s just like they win and we lose. There’s that whole snarl of arguments in there.

One of the things that DeepSeek has not revealed is the data that they train on. You don’t get a copy of the data when you download DeepSeek. You get what are called the weights, and so you get the neural network that results from training on the material that they trained on.

From that, it’s very difficult or impossible to look inside and derive the training data. By the way, neither Anthropic nor OpenAI reveals the data that they train on.

There’s intense speculation in the field as to what’s in the OpenAI training data and what’s not. They consider it a proprietary secret. They don’t release that.

The Chinese DeepSeek may or may not be doing things differently from these companies. They may or may not have a different approach to copyright. By the way, maybe this is all a moot point. Maybe they’re all using the same data, and DeepSeek just figured out a better way to encode the algorithm.

We don’t know. We don’t actually know what the OpenAI and Anthropic algorithms are because they’re not open source.

We don't know how much better or worse they are than the DeepSeek algorithms that have been made public.

Do you think that some of the closed-source models—OpenAI, Anthropic, and any others that enter the fray—end up looking more like Apple relative to Google's Android or something? It's very hard to know what's going on inside Apple. It's an incredibly well-protected IP company.

Embedded in this question is what you personally are rooting for: the set of outcomes that you think are best for America and for the world vis-à-vis AI.

I'm in favor of maximum competition. That fits with the theme of being a VC. If you're a company founder—if I'm running an AI company—I need to have a very specific strategy that has pros and cons and trade-offs. I need to think hard about that.

As a VC, I don't need to do that. I can make multiple bets that have contradictory theses. This is a little Peter Thiel thing, right: determinate optimism versus indeterminate optimism. A company founder CEO has to be a determinate optimist. They have to have a plan, and they have to make the hard trade-offs to be able to succeed at that plan.

A VC is an indeterminate optimist. We can fund 100 different companies with 100 different plans and mutually conflicting assumptions. The nature of my job is that I don't have to make the call that you just described. That makes it easy for me to make a philosophical argument that I really deeply agree with personally, which is that I'm in favor of maximum competition.

I'm in favor of the free market. I'm in favor of maximum competition, maximum freedom, and essentially, if you think about it one level down, a maximum rate of evolution: being able to have as many smart people as possible come up with as many different approaches as possible, run them against each other in the free market, and see what happens.

Specifically for AI, what that means is I'm in favor of the big labs running as fast as they can. I'm 100% supportive of OpenAI and Anthropic doing whatever they want, bringing whatever products to market they want, and running as hard as they want. As long as they're not getting preferential policy treatment, preferential subsidies, or preferential support from the government, they should be able to do whatever it is that they do.

As a company, I'm obviously in favor of startups. I want lots of startups to try lots of different things, and we are, of course, very active in funding AI startups of every shape, size, and description. I want them to be able to run.

Then I want open source to be able to run. I want open source to be able to run in part because I think it's good if stuff shows up in open source, even if it means that there are some business models for companies that don't work. The benefit to the world and to the industry as a whole is so big that we'll find other ways to make money.

AI will just be a lot more common and a lot cheaper and a lot easier, and I think that would be a great outcome. The other really critical thing about open source, and the reason we all need to defend and protect open source, is that without open source, everything just becomes a black box.

Without open source, everything becomes a black box owned and controlled by a small handful of companies that end up essentially colluding with the government, which we could talk about. You need open source to be able to look inside the box and see what's happening, all this stuff.

You also need open source for academic research. You need open source, therefore, for teaching. One of the issues with AI prior to the open-source models, going back 2 years when there were basically no open-source LLMs, was that a crisis was brewing in the universities and the educational system.

University researchers at places like Stanford, MIT, and Berkeley didn't have the money to be able to buy $1 billion worth of NVIDIA chips to actually be in the game. If you talked to computer science professors at the top universities 2 years ago, they were very worried.

The first worry was, “My university's not going to have enough money to be able to stay in the game and do anything relevant in AI anymore.” The other worry was, “All the universities together are not going to have enough money to be able to do anything and stay in the game,” because nobody can keep up with the fundraising of these giant companies.

Open source puts the universities back in the game. It means that if I'm a professor at Stanford, MIT, Berkeley, or any state school—University of Washington, whatever—I can now teach how this stuff works. I can teach using the Llama code, the Mistral code, or the DeepSeek code.

I can do research on that. I can actually continue to make breakthroughs. I can publish my research so people can actually see what's happening. Every new generation of kids coming along who shows up for a freshman computer science course is going to learn how to do this now in a way that they wouldn't if this were a black box.

We need open source in the same way we need freedom of speech, academic freedom, and freedom to research things. My model is basically: let the big companies, the small companies, and open source run and compete against each other.

That's what happened in the computer industry. It worked really well. That's what happened in the internet industry. It worked really well. I believe that's what's going to happen in AI. I think it's going to work really well. All of those can work.

Is there a limit to wanting maximum evolution speed and maximum competition? Perhaps if I said, “We know that the best things are going to come out of China for a set of reasons, because they're willing to do things that we don't let our companies do in the U.S.,” would there be a version of this where you said, “Yes, I want maximum evolution and competition, but the national interest at some level supersedes the desire for maximum evolution and speed of development”?

5. The AI National Interest

That argument is a very real argument. It gets deployed frequently. It is being very actively deployed in the AI space.

As we sit here today, there are actually 2 things. One is that there are currently existing sanctions that prohibit Western and American companies from selling leading-edge AI chips to China. It's not legal today, for example, for NVIDIA to sell its leading-edge AI chips to China.

We actually live in a world in which that decision has been made and that policy has been implemented. The Biden administration had actually put out an executive order—which I think has now been revoked—that was going to apply that same kind of restriction, basically a sanctions process, to software.

This is a very live argument, and there's going to be another round of these arguments in D.C. as a consequence of the DeepSeek thing. Those conversations are underway.

What you have there is a classic thing that you have when you get into policy disputes. You have the rational version of that conversation, which is, “What's in the national interest from a theoretical standpoint?” Then you have the political version of the conversation, which is, “What is the political process actually going to do with the rational argument?”

Let me just say, we all, I think, have a lot of experience watching when a rational argument encounters the political process. It's usually not the rational argument that wins. What you get out of the other side of the machine is not what you went in thinking you were going to get.

There's a third factor that we always need to talk about, which is the corrupting influence of especially big companies. If you're a big company and you're threatened by what's happening in China or what's happening in open source, of course you're going to try to weaponize the U.S. government to protect you.

Maybe that's in the national interest and maybe it's not, but you're for sure going to push for that whether or not it's in the national interest. That's what makes the conversation complicated.

Let's just talk about the chip embargo for a second. You can't sell leading-edge AI chips to China, so that leads to a couple of things.

One is that it for sure sets them back in certain ways. There are certain things that they're not going to be able to do, and maybe that's good because you've decided that's in the national interest. But let's just say there are 3 other interesting consequences of that.

Consequence number 1 is that you have now given Chinese companies an incredible motivation to figure out how to do things on cheaper chips. That's a big part of the DeepSeek breakthrough: they figured out how to use the cheaper chips that are legal in order to do things that require the larger chips for American companies.

That's a big part of the news on DeepSeek, and it's one of the reasons why it's so much cheaper. One of the reasons you can run it on $6,000 worth of hardware is because they put a lot of time and energy into optimizing the code so that it will run efficiently on the cheaper chips that are not under the sanctions.

You force an evolutionary response. That's response number 1. That may have already backfired in a significant way.

Consequence number 2 is that you are incentivizing the Chinese government and Chinese private sector to develop a parallel chip industry. If they know that they can't get American chips, then they are going to develop their own.

They're doing this right now. They have a whole national program to build up their own chip industry so that they're not dependent on American chips. In the counterfactual, maybe they would have bought American chips.

Now they’re going to figure out how to make their own. Maybe it will take them 5 years to be able to make their own. But once they get into a position where they can make their own, then we have a direct competitor on the global market that we would not have had if we had just sold them the chips. And by the way, at that point, we’re in no control over their chips. They’re in total control. They can backdoor them. They can sell them at below cost. They can do whatever they want.

By the way, the other twist on that is Taiwan. China has wanted to reunify with Taiwan for its own reasons for a very long time. But if you’re the CCP right now, Taiwan is a very tempting target, in part because you already want it, but in part because it has these magic fabs. If you can’t buy chips that come off those fabs, but you could seize the island and take the fabs, right, that would be a double reason to invade Taiwan. And so it’s possible these sanctions are going to accelerate the timeline of China doing a military invasion of Taiwan. So that gets complicated. That gets very hairy. The consequences of this get very tricky.

It’s very hard to calibrate these things. And again, the main thing I could just say here is the issues are real, the tensions are real, and the trade-offs are real. There are rational arguments that need to be discussed and argued out. But we need to actually do that rationally, and those of us participating in the system, watching from afar, and voting need to basically keep our eye on this. We need to make sure that the political process isn’t taking rational arguments and twisting them into something that’s either going to backfire or end up just serving a small set of private interests.

This is always the case: these outcomes are so incredibly complicated, and unintended consequences happen everywhere. I’m curious for a little extra thought on the software side. Everyone’s aware of TikTok and whether or not that’s going to persist as an app that our kids can use here in the U.S., but DeepSeek is another great example, where it’s at the top of the App Store and lots of American people are downloading it. I’m curious what you personally are rooting for in terms of our treatment of software provided by China, which tells the CCP, or whomever, tons of information about American users and their behavior, and how you think about that through the practical and ideological lenses.

So I’m a single-celled organism on this question. I have a single function: I want America to win. I want America to win, I want America to win. That’s the only thing I care about. And the reason for that is not because I want America to dominate the world, but because I think the world’s a much better place if America wins, because the American values of freedom and democracy are superior values to, let’s say, totalitarian authoritarian values.

Specifically, though, I don’t want America to win a war. I don’t want to have a war. I don’t think we should go to war with China. I hope we never go to war with China. That would be very destructive. I would like to see America win this Cold War 2.0 the way we won the first Cold War.

So basically, what happened in the first Cold War was the Soviets basically decided by 1989 that the Cold War was unwinnable. They were not going to be able to win. Their economy was not as effective. They were falling behind in technology. They were entering various kinds of social and economic crises as a result of their dictatorial approach to running countries. They had become very corrupt. Their system was not actually working.

Contrary to the American economists at the time, the communist system was not actually working. It wasn’t performing. This is the famous story: if you go to 2 grocery stores, a grocery store in Moscow and a grocery store in New Jersey, you’re just blown away by how much better the grocery store is in New Jersey. You just want it to become so obvious that the correct thing for the Soviet Union in 1989 was simply to say, “We’re done. We’re not fighting this anymore. We’re done. We’re going to stop threatening you. We’re going to stop competing with you. We’re going to stop all this stuff. We’re just going to be a country. We’re going to be a normal country, and we’re going to have normal relationships,” and the Russian people are much better off as a consequence.

I would like to see the CCP and the Chinese people at some point basically just decide, “Look, there’s no point in having this fight. We should just be partners. We should just be partners with America. We should be trading partners. We should have peaceful coexistence, and we should love each other, and it’s all great. We just don’t need to have this fight.”

I believe that the way to win this fight is not to get ultra-aggressive. The way to win this fight is to be strong so that we know that we can obviously defend ourselves if push comes to shove. But fundamentally, we need to double down on our values and double down on our values of openness, freedom, democracy, and peace. In that sense, we need to be strong but not threatening.

So it’s like, yes, if you fight with us, we’re going to win the hard way. But what we’d like to do is just have peace and understanding and trade and freedom. We would like the Chinese people to flourish in the exact same way the American people flourish, and everybody builds a better world.

And this is a big part of my philosophical discussion I get into with people in D.C., which is the impulse that people in D.C. can get into where they’re like, “We need to lock things down, control things, ban things, prohibit things. We need to do all these things, deny things to other countries.” It’s like, okay, but are we sabotaging ourselves? Are we betraying our values? Are we sabotaging our ultimate geopolitical strategic position and the benefits of being America, of being the beacon of freedom and hope and democracy in the world?

In other words, are we going to win by becoming more like China or more like the Soviet Union, or are we going to win by becoming more like America? And so I’m kind of always pushing on the side of, no, let’s be more like America. Let’s be more free. Let’s be more open. Strong, but more friendly, more collaborative, more cooperative, and essentially seduce the world into being on our side as opposed to restricting and controlling and overtly threatening.

How do you think all of this will affect capital allocation? I’m most curious about your firm and how you think it’ll affect Andreessen Horowitz 5 years hence or something like that. If I think about investment firms as a bundle of the ability to form and raise capital, the ability to do great analytical work, and the ability to judge people, especially in the earlier stages, how do you think that function—how capital gets allocated to entrepreneurs who are building the next great companies—will change as a result of o7 coming out and having this incredible reasoning ability and analytical ability? What will be most different about capital allocation and investing because of all of this?

6. AI Reshapes Venture Capital

Yeah, so the analytical component hopefully will change dramatically. One assumes that the world’s best investment companies are going to be very good at harnessing this technology for the use of the kind of analysis that they do, that we do. Now, having said that, the whole thing where the shoemaker’s son has no shoes, one might say that maybe the venture firms that are the most aggressive at investing in AI might be among the less aggressive at actually figuring out how to use it.

We have a bunch of internal efforts underway that I’m super excited about, but firms like ours need to be on the ball here, so we need to actually do it. Some of that work is happening, yet probably not across the industry, probably not enough.

Having said that, there is another side to it. For late-stage investing or public-market investing, a lot of people you talk to have a very analytical lens. There are even great investors—I think it’s Warren Buffett. I think he never does management meetings. I don’t know if this is true or not, but the story I’ve always heard is Warren doesn’t want to meet with CEOs.

He wants the ham-sandwich companies.

Yeah, he wants the ham-sandwich companies, and I think he’s also a little bit worried that he’s going to get seduced by a good story. Well, a lot of CEOs are very charming. I always describe them as having great hair and great teeth. Their shoes are shiny, their suits are impeccable, and they’re really good at selling. Among the things they’re really good at selling is their stock.

And so if you’re Buffett, you sit in Omaha, and what you do is you read the annual reports. The companies put everything in the annual reports, and they’re constrained by federal law to make sure that it’s true, and so that’s how you analyze.

Should o1 or o3 or o7 or R4, or whatever the reasoning models are, be better at analyzing annual reports than at least most investors doing it by hand? Yeah, probably. As you know, investing is an arms race like everything else. If it works for one person, it’ll work for everybody. It’ll be an arbitrage for a little while, and then it’ll close, and it’ll become standard.

And so I would expect the investment-management industry will adopt this technology in that way. This will become a standard way to operate. I do think it gets a little bit different, especially earlier in the process for early-stage venture.

What I’m about to say may just be wishful thinking on my part. I might be the last Japanese soldier on the remote island in 1948. By saying what I’m about to say, I’m going to tempt fate. But I’m going to say, look, so much of what we do on the early side in the first 5 years is really very deep evaluation of individual people, and then it’s working with those people in very deep partnership.

And this is one of the reasons, by the way, that venture doesn't scale well. Venture doesn't scale well geographically. The geographic scale experiments tend not to work, and the reason is that you end up having to be face-to-face with the people for a long time, both during the evaluation process and during the building process. Because in the first 5 years, these companies generally aren't on autopilot. You actually work with them a lot to help make sure that they do the things that they're going to need to succeed.

There's a part of this that is very, very deep interpersonal relationships, conversations, interactions, and coaching. By the way, we learn from them, and they learn from us. It's a lot of back and forth. We don't come in with all the answers, but we have one lens because we see a panorama. They have another lens because they're in the specific details a lot more than we are. And so there's tremendous interpersonal interaction that happens.

Tyler Cohen talks about this. I think he calls it project picking. Certainly, talent scouting would be another version of this, which is basically this: If you look back over hundreds of years, for any new area of human endeavor, you almost always have this thing where you have very idiosyncratic people who are trying to do something new, and then there's some professional support layer of people who fund and support them.

For the music industry, that's David Geffen finding all the early folk artists and turning them into rock stars. Or it's David O. Selznick finding the early movie actors and turning them into movie stars. Or it's the guys sitting in a café, a tavern in Maine, 500 years ago, figuring out which whaling captains are going to be able to go get the whale. Or it's Queen Isabella getting the pitch from Christopher Columbus in the royal chambers and saying, “Yeah, that sounds plausible. Why not?”

There's this alchemy that has developed over time between the people who do the new thing and then the people who sort of enable, support, and fund those people. Let's just say there's no guarantee that this continues, but that's a 400- or 500-year endeavor. Honestly, it's probably thousands of years old. You probably had tribal chieftains 2,000 or 3,000 years ago sitting around a fire, and the young warrior would come up and say, “I want to go take a hunting party into this other thing, and I want to see if there's better game over there,” and the chief's sitting around the fire trying to figure out whether to say yes or no.

So there's something very human about that. My guess would be that that continues. Having said that, if I meet the algorithm that can do that better than I can, I will instantly retire. We'll see what happens.

I mean, you're building one of the biggest firms in this part of the capital allocation universe. How, if at all, have you adjusted your plans and strategy for building the business in the face of this technology, down to the practical “Now we're doing more of this or less of this”? How have you adjusted the direction of the ship based on this new technology?

A big part of running a venture firm, in our view, is that there's a set of values you need to have and a set of behaviors that are what we call timeless. Respect for the entrepreneur, for example. You need to have tremendous respect for the entrepreneur and the journey that they're going on. You need to be deep in the details to really understand what they're doing. You don't do drive-bys. You're building deep relationships. You're going to work with people for a long time.

By the way, the companies are going to take a long time. We don't believe in the overnight success. Most of the great companies get built over 10 years, 20 years, or 30 years. NVIDIA is a great example. NVIDIA's, I think, coming up on its 40th anniversary, and I think one of the original VCs in NVIDIA is actually still on the board. That's a great example of one of these long-term builds.

So there's this core set of beliefs, views, and behaviors that we're not changing at all that kind of have to do with that. Another is the face-to-face thing. You have to be face-to-face with these companies. You do these things remotely. That's on the one hand, but on the other hand, you need to be very up to the minute because the technology's changed so fast, the business model's changed so fast, and the competitive dynamic's changed so fast.

If anything, the environment's getting more complicated because you've got many countries, and now you've got all these political issues that also make things more complicated. We never really worried about the political system interfering with bringing pressure to bear on the things that we invested in up until about 8 years ago, and then it really intensified about 5 years ago. But for the preceding 10 years of us as a firm and the preceding 60 years of venture capital, it was never a big deal. But now it is.

And so now we need to adapt, and we need to be involved in politics in a way that we weren't before. Or now we need to adapt and figure out maybe AI companies are going to be fundamentally different. Maybe they're going to be structured in a totally different way. Or, to your point, maybe software companies are going to work totally differently.

To give you an example of the question we ask a lot right now, it's, “What does the org chart look like for a company that actually fully uses AI? Is it similar, or is it actually very different?” And there's no single answer to that, but we're thinking hard about that. So that side of our brain basically says you need to be up to the minute, and you need to be extremely fluid in reacting to new information.

And then I think the delicate dance that we do every day is to try to figure out what's timeless and then what's up to the minute. That's a big part, conceptually, of how I think about the firm: We need to navigate through that and make sure that we know which is which.

It's so interesting to me that your firm, which obviously has a huge footprint now, is similar in some senses to the KKRs or Blackstones of the world, where you and Ben were founders—seasoned founders—when you started this firm. And similarly with Blackstone, I think Schwarzman had never really made an investment before starting Blackstone, and look what it's become.

It seems like the founder approach—the previous founder approach—to building asset-management and investing firms is that they build these really big, ubiquitous platforms. You've got verticals in most of the major, exciting technology frontiers. Do you think there's some credence to that, that the very best capital-allocation platforms have been and will be started by founders more so than by investors?

Yeah, a couple of things. One is, I think the observation makes a lot of sense. The way that people in the business talk about it is basically that there's a lot of investment operations that are—usually, the term you hear is “partnership,” right?

A lot of venture firms have historically worked like this. It's just a partnership. It's like a small tribe of people sitting in a room together, trying to bounce ideas off of each other, and then they make the investments. By the way, they don't have a balance sheet. It's a private partnership. They pay out the money in the form of compensation at the end of every year. That's traditional venture capital: a traditional venture capital model with 6 GPs sitting around a table doing that. They have their assistants. They had a couple of associates. But the point is, it's completely based on the people.

And by the way, it actually turns out that in most cases, what you discover is that people don't actually like each other that much. Mad Men did a good job of this. Remember in Mad Men, when the partners went off and started their own company in the 3rd or 4th season? They actually didn't like each other. They knew they had to come together to start a firm, and that's how a lot of these firms operate.

And so they're like these loosely affiliated tribes, and then you get this phenomenon they call “you eat what you kill.” Each person wants to make the most of the money off the things that they do, and they don't want freeloaders. But it's a partnership, and it's a vote to try to change anything, and so it's hard to fire people.

And so anyway, it's a private partnership, and it's fine for what it is. But then what you see with those is that they have a hard time sustaining. There's no franchise value. There's no underlying enterprise value. It's not a business.

What you see with those is that when the original partners at a firm like that are ready to retire or do something different, they hand it off to a new generation. Most of the time, the new generation can't keep it going. And most firms in that model, I think, are now kind of phasing out for that reason.

But even if they can keep it going, there's no underlying asset value. That next generation is just going to have to hand it off to the third generation, and it's probably going to fail in the third generation. Then it's going to be on Wikipedia someday as, “Yeah, that firm existed at one point, and then it went away, and other firms took over”—ships passing in the night.

So that's the standard way to do it. And, by the way, if you're trained as a classical investor, you've been trained how to do the investment part, but you've never been trained in business building, to your point, and you've never built a business. And so it's not natural for you to build a business. You don't have the skill set or experience, and so you just don't do it.

Many investors have made tons of money as investors running in that model for a long time, so it can work really well. The other way is to build a company, build a business, and build something with enduring franchise value in it.

You alluded to firms like Blackstone and KKR, which are huge public companies—Apollo, these huge firms. You probably know the original banks actually were all private partnerships. Goldman Sachs and J.P. Morgan 100 years ago looked like little venture capital firms, much more than they look like today, but then their leaders over time turned them into these huge franchises, and they're also big public companies.

So that's the other thing to do: build a franchise. Now, to do that, you need a theory as to why a franchise should exist. You need a conceptual theory as to why it makes sense to do that. Then you need the business skills.

At that point, you're running a business, and it's like running any other business: “Okay, I've got a company. It's got an operating model. It's got an operational tempo. It's got management capabilities. It has staff. It has multiple layers. It has training programs. It has performance management. It professionalizes its operations. It has division of labor internally with specialization.” Then you think in terms of scaling, and over time, you think in terms of underlying asset value, where the thing has a value that's not just the people who happen to be there at the moment. It's not like we're chomping at the bit to take it public, but a big part of what we've been trying to do is build something that has that kind of enduring aspect to it.

Are there aspects of the firm that you hope are new and different in 10 years that don't currently exist? And are there any non-compromisables, ways that you hope the firm never evolves to look more like traditional large-scale asset managers?

We evolve very rapidly in what we invest in, and so what the companies are, what they do, what the models are, what the founder backgrounds are—that stuff changes all the time. I'll give you an example. For 60 years of venture capital, the one thing you never did—everybody knew the one thing you never do—was you never back a researcher. You never back a PhD to start a company and do research. They'll just do research, burn the money out, and you'll have nothing at the end.

Now, many of the best AI companies were founded by researchers. That's one of those things where it turned out that it was not a timeless value. That needed to be an up-to-the-minute thing that you needed to adapt to. We need to be very fluid on that.

As a consequence of that, what those companies need to succeed, what they need from us, and the kinds of help that they need from us also change. The most significant change at our firm—I mentioned it before, but we now have a big political operation. Even by general business standards, we have one of the largest and now most sophisticated political operations, I think, in the business world. Four years ago, we had nothing. We had zero in politics. So that's a function that we didn't imagine we would need.

I'm sure that in 10 years we will both be investing in things I can't even conceive of today, and we will have operations that I can't even conceive of today. So I'm sure that will all change. We're totally open to change in all those dimensions.

There's a bunch of timeless values. I would hope that the firm values are going to be the same in 10 years because I think those are pretty well thought through. But the other thing that I'm always telling our people internally and always telling our LPs is that we are not trying to build for scale in order to be a large-scale asset manager.

The majority of the investment industry is asset gathering. A firm reaches a point where it just decides to go for size and scale—from tens to hundreds of billions of dollars to trillions of dollars—and it's not going for great returns anymore. It's just going for scale and mass. The accusation of firms that do that is that they're trying to harvest the fees more than they're trying to outperform on the investments.

That's not what we're trying to do. We're not going for scale for scale's sake. When we go for scale, it's because we think it's necessary to support the kinds of companies that we want to help our founders build.

But the way I describe it is, the core of the firm is always and will always be early-stage venture. So no matter how big we get as we raise these larger growth funds to be able to write bigger checks—some of the AI companies need very large amounts of money—we didn't start with a growth fund, and now we have a growth fund. But the point is, the core of what we do is always going to be early-stage venture, and that confuses people a bit.

I would say people get confused on this because, from the outside, it's like, “You guys are managing so much money. Why, as a founder of an early-stage startup, would I believe that you would want to spend time with me, because it's not worth your time? You, a16z, are going to invest $5 million in my Series A, but you have these other investments that you've invested $500 million in. Why are you going to spend time with me?” And the reason is because the core of the business is early-stage venture.

By the way, the return opportunity off that early-stage investment is as big as the return opportunity from the later-stage companies, which is a characteristic of startups. So, Number 1, financially, it actually makes sense for us to spend time with the early stage.

But the second is all of our knowledge and all of our relationships and everything that makes us special as a firm: all the deep insight and all the people that we know and who trust us at that early stage. What I always tell people is, “Look, if push comes to shove and the world goes sideways and we need to sacrifice things, the thing that will never get sacrificed is the early-stage venture business. That will always be the core of it.”

A big part of what I try to do with that is then have a lot of my time free. I think generally our founders actually get fairly surprised by this: I have a lot of time free in order to work with the early-stage founders. Number 1, it's super fun, but also, you learn the most.

If you think about the changing nature of power structures in the world, we talked a lot earlier about America and China as Cold War 2.0. I've always seen you tweeting about tons of books studying this specific topic, transitions in power through world history. Where else do you have your eye today on power centers that are changing, either gaining or losing power, that interest you most?

7. The Rise Of Counter Elites

The Machiavellians—I'm sure you've probably had a dozen people on your show who have recommended it. It's one of the great books of the 20th century, and it goes through this theory of political power and social and cultural power. One of the key things in that book that I see everywhere right now is this idea of elites and counter-elites.

The idea goes like this. Basically, democracy per se is a myth. You never have an actual fully democratic society. By the way, the US, of course, is not a democracy. It's a republic.

Experiments in direct democracy historically have worked very badly. The systems that work, even the democratic systems that work, tend to be republican in nature—small-r republican in nature. They tend to have a parliament or a House and Senate or something. They have some sort of representative body.

The reason for that is a phenomenon described in that book called the Iron Law of Oligarchy, which is basically the following: The problem with so-called “direct democracy” is that the masses can't organize. You can't actually get 350 million people to organize in anything. It's too many people.

So what you have in basically every political system in human history is a small organized elite that governs a large disorganized mass. You had that in the original hunter-gatherer tribes all the way up through America and every other political system in the modern era. By the way, this was the Greeks, and it was the Romans, and it was every other political society that ever worked. Every empire in history, every country in history. So, a small organized elite governing a large unorganized mass.

That's a relationship that's fraught with peril because the unorganized masses will go along with the elite for a while, but not necessarily forever. If the elite becomes abusive toward the masses, the masses greatly outnumber the elites, and at some point they're going to show up with pitchforks and torches. So there is some tension there, and a lot of revolutions occur when the masses decide that the elites are no longer representing them properly.

Our society is no different from any other society: We have large disorganized masses, and we have a very small organized elite. We have 2 elites. Our political system, which our founders set up, has our Democratic elites and our Republican elites, with quite a bit of overlap between them. Some people actually call that the uniparty. Maybe those elites have more in common with each other than they do with any of the people.

Those elites cruise along. We had an establishment Republican elite for a long time, with its policies that culminated in the Bushes. We had a Democratic elite with its policies that culminated with Obama.

In the last decade, on both sides in the US, there's been a revolt within the elites, and this is actually the key point in The Machiavellians: The way change happens usually is not the masses activating against the elites directly. What happens is the emergence of a new counter-elite. You'll have a new counter-elite that will compete to take over from the current elite.

My read of current affairs broadly throughout the world is that, generally, the elites that run the world are being found to have done a bad job. We can talk about why in a second, but generally speaking, if you look at approval ratings of political leaders and approval ratings of institutions, everything is crashing.

The uniform thing that's happening in the world—political people call this anti-incumbency. Basically, if you're an incumbent institution, if you're an incumbent newspaper, if you're an incumbent TV network, if you're an incumbent university, if you're an incumbent government, generally speaking, your poll ratings are a disaster. That's the people basically saying, “The elites in charge are failing us.”

And then you have the emergence of these counter-elites who are coming along and saying, “Oh, I know. I have a better way to represent the masses, and I have a better way to take over, and my new counter-elite movement should take over from the elite movement.” In the Democratic Party, this was Bernie Sanders in 2016. It’s AOC. It’s that whole wave. The Republican Party, obviously, is Trump, and it’s the MAGA movement and everything that represents.

But by the way, this exact same dynamic is playing out in the UK. The Tories have collapsed, and now you’ve got this Reform Party with Nigel Farage that is very threatening. You had Jeremy Corbyn, who was a counter-elite coming in from the left. You have that in Germany. Actually, this week in Germany, there’s this very dramatic thing happening, which is that this quote-unquote “far-right party,” AfD, is rising very fast, and there’s this leader, Alice Weidel.

This is the first week in German political history in 50 years or whatever that the CDU in Germany actually partnered with AfD on something. All of a sudden, AfD is a viable competitor. They’re the counter-elite trying to take over the right of the German political system. So basically, everywhere in the world you go, you’ve got a counter-elite showing up and saying, “Oh, I can do this better.”

It’s a fight between elites. It’s a fight between the elite and the counter-elite. It’s a fight that the masses are aware of and they’re watching. In democratic societies, they’re ultimately going to decide, because they’re going to decide who they vote for. This was when Republican voters decided they were going to vote for Trump and not Jeb Bush. That was the counter-elite beating the elite.

This actually goes to the critiques of Trump, which is so interesting. Trump gets criticized a lot by the existing elites as, like, “Oh, he’s not actually a man of the people. He’s a super-rich billionaire who lives in a golden penthouse and gets driven around everywhere in a limo.” If you’re a rural farmer in Kentucky or Wisconsin, you shouldn’t think this is one of your people.

The point never was that Trump is a man of the people. The point was that Trump is a counter-elite who’s able to represent the people better. That’s the whole basis of his movement. Anyway, that’s the general pattern, and I just think you see that everywhere in our society.

And then, look, you’re an example of this with what’s happening in the media. Everything I just described is exactly what’s happened in the media. You had elite media for 50 years, and it was network news and cable news and newspapers and these prestige magazines. Now you’ve got the counter-elite, and the counter-elite is you and Rogan and on and on.

By the way, you look at the numbers, and it’s very clear where the people are going. The people, the viewers, the masses, the readers are fleeing the old, and they’re going to the new. The incumbent elites are absolutely furious about it. They’re furiously writing all these hit pieces about how you guys are all a bunch of white supremacists and the whole thing is horrible, right?

And it’s like, yeah, it’s the way of the world. We’re in the middle of this. I don’t even know if “just transition” is the right term. It’s a pitch battle, basically, between old elites and new elites.

What were the original seeds of demise for the previous generation of elites that led to all those 11% approval ratings? What would you attribute most of that to?

There are really 2 theories. There’s the theory that those approval ratings are wrong, and there’s the theory that those approval ratings are right. By wrong, I mean they’re being measured correctly, but the people are wrong. People are giving the wrong answer.

And so if you’re running CNN or Harvard or whatever and your approval rating is coming in at 11%—and by the way, for your listeners, Gallup has been doing this incredible survey panel for 50 years, which is trust in institutions—you can just Google “2024 Gallup trust in institutions,” and you get these spectacular charts. What you find is basically trust in institutions peaked in the late ’60s and early ’70s and has been falling off a cliff ever since.

This phenomenon, by the way, predates the internet, interestingly. It gets blamed a lot on the internet, but it predates the internet, so it’s something that basically started developing in the ’70s and has been accelerating since. Those ratings have fallen off a cliff dramatically since 2020. They slide like this, and then after 2020, they just plummet.

TV network news—you know the number. It’s tiny. It’s a single-digit percentage. People are just completely done with it. They don’t believe what’s on it anymore at all. Viewership is collapsing in the same way.

So one theory, if you’re running NBC News, CNN, Harvard, your theory is, “Oh, the people are wrong. The people have been misinformed. They’ve been lied to. They’ve been fed misinformation.” This is why the whole misinformation thing became such a big meme. There’s a Marxist concept called false consciousness, which basically says that the people are incorrect.

The people have been lied to by malicious actors, by populists and demagogues, and it’s just a matter of time until we can explain to the people that they’ve been lied to, and they’re going to come around, and they’re going to believe us again. So that’s one theory.

The other theory is that the elites have become rotten. They have become rotten and dysfunctional and corrupt, and they’re not delivering. On that theory, the numbers, the ratings are correct. The collapse in approval is correct because every time you look at Congress, they’re turbo-shotgunning money out the door at all kinds of crazy stuff. It’s taken out of your pocket as a taxpayer with no regard at all for you.

If you go watch CNN or NBC News, they’re just lying to you about a thousand different things all the time. If you go to Harvard, they’re teaching you race communism, that America’s evil, and crazy, crazy things. In that theory, the people are correct. The people are onto these elites.

These elites basically have been in power too long. They’ve had too much power. They haven’t had enough scrutiny. They haven’t been subject to enough competitive pressure, and they’ve rotted in place, and they’re just not delivering anymore. The reality is probably some of each.

It’s very easy for the next rabble-rouser to show up and just start hurling stones at whoever’s in power and saying whatever. I don’t want to pick on people, but if you’re somebody who doesn’t have political power today but you want it, the easiest thing to do is show up and start yelling about how the current elites are corrupt.

Maybe it’s somewhat correct. Demagoguery has something to do with this, or whatever, and misinformation, but I think an awful lot of it is that the elites are rotten. My version of that is very straightforward, which is that Burnham talks about this in the book. There’s this concept called the circulation of elites.

One of the things he says is, for an elite to actually stay healthy and real and productive and not rot in place, it needs constant infusions of new talent, and it does that through a process of circulation of elites. What it does is identify young, promising people and invite them to join the elite.

It does that for 2 reasons. One is so it can refresh itself, and then the other is that those are the people who would be the most likely to become the counter-elite. So it’s also a way to head off future competition.

This has been my experience basically since I was 22: “Oh, hey, Marc, we would love for you to come to Davos. We would love for you to come to Aspen. We would love to have you come to this big conference in New York. We would love to invite you to the dinner parties. We would love to have you come hang out with reporters at The New York Times.”

For 25 years, this is what I did: “Oh, that sounds great. These are the best people in the world. They’re in charge of everything. They’ve got the best degrees. They went to the best schools. They’re in all the positions of power. They love me. They think I’m great. They keep complimenting me, and I’m in these rooms with these important people, and they’re taking me seriously.”

It’s just this ego bath that’s incredible. “Wow, I’m like a kid from the cornfields of Wisconsin. I’m like, I’ve arrived, and I’m in the elite.” All I have to do to stay in the elite is not argue with anything. All I have to do is just agree with whatever’s in The New York Times and whatever’s being said at Davos, vote for the candidates that you’re supposed to vote for, donate to the ones you’re supposed to donate to, and never, ever, ever deviate off the track.

Then you just become part of the elite. I have lots of contemporaries who have done that. Some of them are the world’s largest Democratic donors now, where they’re fully minted, and they’re in there, and they’re having a great time, and they think it’s all incredible and wonderful.

Some people are fine going along with that, and maybe that’s the right thing to do. Then some of us reach a point where we look around. This is the J.D. Vance story. He tells a very similar story. Grew up in rural Kentucky, whatever, Ohio, Appalachia. He ends up at Yale.

He ends up being invited into the inner circle of all these places, and he finally looks around, and he's just like, “Wow, these people are not at all what I thought they were. This is horrible.” These people are self-interested, corrupt, and they're lying about everything, engaging in speech suppression, incredibly authoritarian, and looting the public treasury.

Oh my God, I was lied to my entire life. These people don't deserve the respect that they have, and maybe there should be a new elite in charge. So that's a lot of the tension that's playing out right now. Yeah, I'm a case study of it myself.

If we take an optimistic lens, your emphasis on early-stage venture means you get to meet all these young, brilliant people who are about to go build the future. Let's take the optimistic lens and say AI has the most positive impact that we can imagine in all the places where we can verify the outcomes. Reasoning becomes so powerful.

What are the other related choke points that get in the way of the sort of explosive technology revolutions that we all want? That could be clinical trials in medicine or something that's just going to go slower than AI, and AI's not going to be the problem. We're going to be bursting at the seams, wanting to make progress, but the world of atoms, the world of regulation, the world of clinical trials, or whatever it might be, are going to become the rate limiters—not intelligence and knowledge. Which of those choke points have you most personally interested in today?

8. The New Technology Bottlenecks

The way I've always thought about technological change is as a set of lines on a graph. It used to be 3 lines; now it's 4 lines. One is the pace of technological change. That's a line where everything generally gets better and better and better, and then every once in a while, you have these discontinuous step functions up where something gets dramatically better, like what happened last week with AI.

Then you've got another line on top of that, which is sociological change, which is basically: When is the world ready for the new thing? Sometimes you get this phenomenon where the new thing actually exists before the world's ready for it, and for some reason, it doesn't take. Then 5 years later or 50 years later, it actually takes, and off and away it goes.

So there's the sociological layer, and then on top of that, there's the financial layer, which is: Are the capital markets willing to fund it, and can it generate a return? The way I think about it is the technology line is a squiggle like this, up and to the right. The social line is this big, sweeping curve as large numbers of people reevaluate what they want. And then the financial line is like an EKG of a heart attack, where the market's going through its patterns of panic and euphoria.

The art of being an entrepreneur or a tech investor is to try to slice across all 3 of those. You're trying to back something where the technology's really ready, society's ready to adopt it, and you can actually get the thing funded or get the thing exited and taken public. So you have to line that up. A lot of what we do in the day job is line up those 3 curves.

The 4th one now, in the last 5 years, is politics. As I said earlier, for a very long time, people in Washington just had an attitude of benign neglect toward the tech industry. They were like, “Yeah, their kid's building fun toys, whatever. It's fine. It doesn't matter.”

And now there's, as you know, intense political scrutiny on almost every aspect of the tech industry, and incredible attempts to control and suppress new technologies. The harshest version of that is in places like the EU, but we've had our own versions of that here. And so all of a sudden, we have this 4th factor now.

In the last 4 years, overwhelmingly, the answer to your question is politics. In the last 4 years, overwhelmingly, the biggest issue that we had was government. That was very bizarre and disconcerting to me when it first started because I wasn't used to it, and I had never viewed us as being involved in politics or being partisan. We were not in Washington trying to curry favor. We were not trying to get subsidies, but we also didn't think we had to do anything to avoid getting stepped on.

And then that just radically changed. That's the single biggest thing. You're probably well aware at this point, or your listeners are well aware, the policy implications on crypto were just devastating for the last 5 years. Fintech was also extremely damaging. Social media got tremendously stomped on by the state, which led to all the censorship stuff.

And then AI—I've spoken elsewhere about what they were basically going to do to AI. For the last 4 years, I've just been in this bizarre situation where it's like my main enemy is my own government, which is very, very strange. And by the way, talk about accelerating my evolution from an elite into a counter-elite. It's like, okay, if they hate me and want to destroy me, it makes the call relatively easy for what I need to do.

How did you most feel that? How did you feel the elite wanting to destroy you? How did that manifest most?

It was sort of coincidental with the national mood shift, probably between—I would describe it as between 2013 and 2017. I grew up politically as a child of the 1990s. I happened to be in business at that point and high-profile, so I knew Clinton and Gore quite well, and I was sort of just a default Clinton-Gore Democrat.

It was great. It was what I call the deal with a capital D: You're a Democrat, but the Democrats are pro-business, pro-tech, and pro-startup. Clinton and Gore love Silicon Valley. They love new technologies. They were always thrilled to see what we were doing. They were incredibly supportive. They would try to help us. If other countries came at us or whatever, they would try to help us and support us.

And yeah, it was great. You could be a pro-business, pro-tech Democrat. It was great. You could make a bunch of money. People would write all these great articles about you, and then you give all the money away, and you're a philanthropist, and it's great. You die, and your obituary says, “He was a great entrepreneur and a great philanthropist,” and everything is wonderful.

That deal collapsed starting in 2013. Specifically, every single part of that deal collapsed in 2013, including when a lot of the political system turned on philanthropy, which I found to be probably the most amazing thing of all. It turns out philanthropy is evil, if you haven't been updated on this yet.

Because the correct role of money is that it should all go to the government, and the government should hand it out. They created a slur around 2013 called “philanthrocapitalism.” Basically, it described rich people who thought they could make private choices about how they gave away money, as opposed to letting the government do it, and that this was evil.

Every aspect of that deal between, let's call it, the tech elites and the Democratic elites broke down. That showed up in a thousand ways, but it showed up as press coverage. The official organs of the elite turned on us, and everything we did was evil. It was actually fairly amazing.

In 2012, social media was considered an absolute, unalloyed good by the mainstream press because it had gotten Obama reelected, and it had been the catalyst for the Arab Spring. Everybody knew that it would only ever get the right political candidate elected, like Obama, and everybody knew that the Arab Spring would result in peace and democracy throughout the Middle East forever.

Then by 2016, the narrative had completely flipped to social media, the internet, and tech destroying democracy, and everything being ruined by it. And so the press coverage—all of that was a canary in the coal mine. Part of it was that the employee base got radicalized. A bunch of the professional investors got radicalized, too.

We had this bizarre situation where you had these big investment managers showing up demanding radical politics in your company, which was completely bananas. At the time, there were a thousand other versions of this, and then ultimately what happened was the government itself showed up.

The bureaucracy under Trump started to do this kind of thing outside of his direct control. But under Biden, it became a concerted campaign of destruction, with an endless barrage of prosecutions, investigations, Wells notices, debanking, censorship, and attacks—a comprehensive attempt to basically destroy entire sectors.

Of course, that's what we ultimately ended up reacting to. My hope is that's over, which is to say the new administration is taking a very different approach and not doing all those things. And then my hope is the next Democratic government basically realizes that attacking tech and attacking startups was actually not necessary and, in fact, was probably counterproductive, because if you drive Elon Musk out of your party, it has consequences.

I talk to a lot of Democrats. We support a lot of Democrats at the firm—a lot of Democratic congressmen and senators—and I talk to them a lot. I'll be out there again next week talking to them.

What they tell me is, “Look, there's a civil war inside the Democratic Party between those of us who think the party should come back to the center and just stop attacking capitalism, attacking business, and attacking tech, and get back to winning elections. And then there's a bunch of us who think that the party actually needs to become more radical. We need to differentiate more from the other side, and we need to become—I don't want to use pejoratives—but more extreme in economic policy, more extreme in tech policy, and more extreme in social policy.”

They're fighting that.

My hope is that they’re going to work their way back to the center so we never have to go through this again, and we can have positive relationships with both sides. But we’ll see what happens.

I’m incredibly interested, as so many are, by the nature and state of the global supply chain. This is indirectly a little bit of another China question, but when you dig into the components of pharmaceuticals, or really the components of lots of things, you see how incredibly interdependent the world is, especially the U.S. and its reliance on places outside of the U.S. for supply chains, generally speaking.

I’m curious how you think and hope this will evolve in the next decade or so. Obviously, there were benefits—we went global for a reason—but I think there are now lots of fragile spaces in supply chains around the world. How do you think about this part of the evolving economy and the economic story? To go back to your point, you want the U.S. to win, so how do you think the U.S. should and might win in supply-chain manufacturing and all these exciting ideas you hear swirling around today?

9. The Global Supply Chain Ladder

Yeah. This is really important, and this is where it’s so much different from the past. This is where our interaction with China is so different from our interaction with the Soviet Union 50 years ago. To your point, it’s the overall level of activity of offshoring. American manufacturing companies never really offshored anything into the Soviet Union, but they offshored a lot of stuff into China, so there’s a lot more activity that’s gone overseas.

The other thing—you know this, but it’s really important—is the complexity of the supply chain. Take the iPhone as the canonical product. There’s a document you can download online that’s probably a little bit dated now, but it goes through the componentry that makes up an iPhone and where that stuff comes from. The one I read a decade ago had parts in the iPhone from 40 different countries. There might be a more recent one, but by the time that iPhone is getting assembled at Foxconn in China, 39 other countries have literally sent stuff in, getting built into subcomponents of subcomponents and then into components.

Cars are the same way, and robots are going to be the same way. Anything sophisticated, anything computerized or mechanical, is going to have that attribute. By the way, it’s actually hard to even get this from the trade numbers, because I believe this to be the case: China gets to take credit for the export value of the completed iPhone in its export numbers, even though the economic value added of what happened in China is in the single-digit percentages.

Because so much of what’s in the iPhone comes from 39 other countries, the analysis you really want to do is what’s called economic value added. You want to basically say, “Okay, of the $1,000 that went into the iPhone, what’s the pie chart of the value of where those things came from in dollars?” The answer is: from all over the world.

This is the problem with a simplistic argument about reshoring or about reversing globalization. We’re not talking about bringing a steel plant back from China. We’re talking about unwinding a supply chain that has 40 countries involved, with things going back and forth all over the place as everything is built up and assembled.

The reason an iPhone costs $1,000 and not $1 billion is because the efficiency gains from that level of economic specialization and trade have been profound. We have a material standard of living today that’s far higher as a consequence of this. Our standard of living in the U.S. is much higher than it was, and billions of people around the world have been brought out of poverty. Economically, that system has done quite well in many ways.

The problem is that it runs up against reality in a bunch of ways. One is national security. This is the problem with the Taiwan chips and the Taiwan fabs. If the U.S. military is going to be running on autonomous, AI-driven drones and self-piloted fighter jets and self-piloted submarines in the future, as opposed to what we have now, we’re going to need those leading-edge chips to power our military. If China seizes Taiwan, it owns the means of production for those chips, and the American military doesn’t get them.

So there are direct supply-chain implications for the military. There’s also the geopolitical issue we talked about before: do countries start to use their ownership of different areas of this more as leverage in geopolitical fights?

Then there’s what we saw under COVID. When the world goes into crisis, there can be a big fight over even things that you would consider relatively prosaic. This is a great example. You remember in the early days of COVID, there was this idea that if you did enough testing, you could do what was called test and trace, so you could isolate COVID clusters before they spread, which is a standard thing that people try to do for infectious disease.

It turned out they couldn’t get the COVID tests out fast enough to do it. It turned out that the plastic tip that goes into the COVID testing thing was made by some factory in China. It’s a piece of plastic, but if you can’t get it, you can’t make the test. It doesn’t even have to be the world’s most sophisticated component that ends up holding you up. It can be relatively simple things if you’re not capable of building them internally, and that could determine which country can respond to a pandemic.

Then you just have the political and economic pressure of it. We all assumed—the American political system assumed—for 30 years that you could just offshore manufacturing out of the U.S., and that the communities that saw all their plants close throughout the Midwest and the South were just going to sit and take it. They were just going to figure out something else to do.

In a lot of the U.S., they never figured out a new thing to do, and it turned out they can still vote. Part of what’s happened in our political system is that they’ve decided they’re just not having it anymore, and they’re going to vote for something different.

People argued about that at the time, but the economic-efficiency argument won and had its benefits. It paid off in some ways, but a lot of people in the country were radicalized. I come from a part of the country in which a lot of people were radicalized by the fact that the government and businesses apparently thought it was fine to just hollow out the economy and send everything offshore.

Even if you’re getting the payoff from the economic efficiency, your political system may not be able to withstand that. You may end up really regretting it. I don’t think there are any easy answers here. Anybody, in my view, who says there’s an easy answer here is wrong. This is complicated.

Probably what’s going to happen is that the world will remain extremely interdependent, and there will be a lot of pressure and a lot of back-and-forth. This whole dynamic that’s playing out with tariffs and trade negotiations will be a constant thing, as it has been forever, and there will be twists and turns along the way.

Fundamentally, the world will stay interconnected in lots of ways, and we’ll muddle through it. The fear is that at some point there’s a war, or an even more severe pandemic, or something like that in which all of this gets stressed to the point where it really breaks hard. I hope that doesn’t happen, but in a way, the more interconnected the world gets, the more resilient it gets, because there are more ways to do things and more ways for people to adapt and for everything to change.

Then, in some ways, the more interconnected the world gets, the more dangerous things get, because if any one part of it breaks, the whole thing breaks. There’s a real push and pull on that.

There’s another lurking area of technology frontiers that I haven’t seen you talk about a ton, which is robotics. Everyone’s very excited by the potential. It’s very easy to imagine a humanoid robot lurking around, doing all sorts of stuff that humans don’t have to do anymore.

There are tons of technological breakthroughs required to make that world a reality. What do you think is going to happen in the world of robotics? What’s being overestimated? What’s being underestimated? How do you think about it?

I would make a list of 4 things: phones, drones, cars, and robots. Basically, this is the ladder that China’s climbing. To your point, this is the ladder of not just products, but entire supply chains.

China became the place where all the phones were assembled and manufactured. As you know, it built up this entire ecosystem in China of thousands and thousands of specialist companies to do phones. This particular environment is called Shenzhen, which is a cluster of thousands and thousands of companies that manufacture all kinds of electronic, hardware, mechanical, and computer-related things.

They did that in phones, and they produced billions of phones a year. That worked, and it’s running at high scale. That supply chain was then leveraged for China to win the drone market—consumer drones like DJI drones.

Basically, what happened was that China won the global drone market. It has a 99-point-something-percent share of the global drone market. By the way, as a consequence of that, it has well over 90% of all the drones used by the American military, which is interesting, because every single one of those drones is potentially backdoored to be a surveillance platform or to be used as a kamikaze weapon. There’s a real issue with this—a great example of the issues that emerge.

But a big part of why China wanted drones, at least up until now, was that they had this entire supply chain that started out building phones, which they adapted, and then they built all this stuff to make drones. A drone, in a lot of ways, is like a flying phone. It has a lot of the same equipment, then it has some new stuff, but they wanted that, at least up until recently.

Now they’re going to cars. The reason they’re going to cars is because a modern self-driving electric car is much more like a rolling laptop on wheels, or a rolling phone on wheels, than it is like an old-fashioned internal-combustion car. Tesla is our example of that in the U.S., where a Tesla is a computer and a lot of batteries wrapped in a frame with some tires.

The great illustration of the change here is if you just visit the service bay at a traditional car dealership versus the service bay at a Tesla dealership. The service bay in a traditional car dealership is oil and grease everywhere, with all kinds of stuff going on. Everybody’s got the overalls, and they’ve got the dirty rag that they’ve been wiping their hands on all day.

You go to a Tesla dealership, and it’s like an operating room. Everything’s clean because it’s not internal combustion. There’s none of this stuff. It’s just a computer.

The Chinese basically are now doing in cars what they did in drones and what they did in phones, which is that they’ve built an entire ecosystem leveraging those other supply chains. They’ve built an entire ecosystem of all the parts needed to build self-driving electric cars, and they’re now bringing those cars to market. Now, all of a sudden, they’re really good.

They’re really good in the same way that Chinese phones are really good and Chinese drones are really good: They’re fully modern, super advanced, super inexpensive, and leading-edge technology. The cars are getting to be really good, and they’re a third or a fourth the price of the equivalent car in the U.S.

The fourth phase is robots. If you have the supply chain for phones, drones, and cars, you have most of what you need to do robots, and that’s the next phase. They’re doing it. We have, obviously, Elon and other companies in the U.S. building humanoid robots, and I hope and expect that they’ll do well, but China is doing that for sure.

The company I’ve been watching most closely is one of their national champions. It’s a company called Unitree. We’re not involved in this, but Unitree sells a robot dog that’s equivalent to the Boston Dynamics robot dog. The Boston Dynamics robot dog costs between $50,000 and $100,000, which is why you don’t see very many of them. The Unitree dog starts at $1,500.

By the way, we have 2 of them, and they’re great. They do backflips, they climb stairs, and they talk to you. They have an LLM built in, and they’ll teach you quantum physics as you’re running around in the yard. It’s great. They have humanoid robots coming out now that are also at much, much lower prices.

They are coming for robots in a major way. Again, this is going to be a real push-pull, because it’s like, all right, if you believe that humanoid robots are going to happen, which I do, and at large scale, and if China’s willing to make them for $10,000 or $20,000, and we can buy a billion of them, and all of a sudden we have robots building our houses, doing lawn care, and doing everything else that you’d want robots to do, waiting on you hand and foot, then it’s great that China’s making them and selling them to you, that they’re super cheap, and that they work really well.

Having said that, if there’s ever a war between the countries, every one of those robots could go rogue and start to attack you. You might want to think about that. It might be important to have robots that are made in the U.S. It might be important that the robots the military uses are made in the U.S.

You might want the robot in your house to be made in the U.S., or the robot that’s taking care of your kid and changing his diaper. Phones and drones are already intense issues, but cars and robots are going to be ultra-intense. It hasn’t quite happened yet because the robot thing hasn’t quite tipped, but I think the robot thing is going to tip in the next few years.

Then I think there’s going to be a giant geopolitical drama that’s going to play out to try to figure out what we should do.

It’s been fascinating to watch the race to build not just the bodies of the robots, but the brains. Companies like Physical Intelligence and others, American-based, are trying to build the datasets that we just don’t have yet. We had the open web to train AIs on.

Are there areas elsewhere, in all the young people you’re seeing and all the companies you’re seeing, that you are incredibly excited about, but you feel like the market has not yet realized what is going on and what is possible?

10. The Biotech Frontier

I guess maybe biotech. The good news is that, in the modern world, there are a lot of people who are into new technology, and there are a lot of people who talk about it. When I was a kid, early-adopter markets were tiny, so the number of people who wanted their first personal computer, whatever, was just a tiny number of people.

Now you’ve got 50 or 100 million early adopters who just want whatever is the new thing, talk about it all the time, and talk about it online. I don’t know that there’s that much of a delay anymore, but probably biotech.

Everything that’s sort of in the cluster of life extension, embryo selection, and potentially the reproductive technologies—getting embryos from stem cells. Couples that can’t have children—you probably know a lot of people like this—people who get to a point where either they had a fertility problem when they were young or they get to an age where they have fertility issues, but they want more kids. Then they’re forced into very difficult choices having to do with IVF or donors of different kinds.

It looks like we’re going to be able to have embryos from stem cells, and so you can have children who are actual biological children much later in life. External gestation is a while away still, but at some point, that might be a big deal.

People talk about the birth rate a lot. Well, if you could continue to have kids into your 60s, and if you could have a dozen kids because you could have external gestation, would more people choose to do that? Maybe yes. So that would be one.

Another might be genetic optimization. One of the endlessly spicy topics is intelligence augmentation. There’s CRISPR. We now have gene-editing technology, so in theory, you can go in and do genetic reprogramming of people, especially at the early stages. The scientists are figuring out the hundreds of genes that correspond to IQ. Should you have the ability to boost IQ? That has all kinds of downstream questions.

So probably it’s in that vanguard, those kinds of vanguard movements. The inflammatory version of this was the guy in China who made the babies who were immune to AIDS. There was a Chinese guy who did a home-brew CRISPR thing, and he created 2 embryos that are now, I think, living children, as far as I know, and used CRISPR to splice in genes that make them immune to HIV.

It became this dramatic firestorm of controversy because CRISPR is not developed well enough yet to do that in a predictable way. There were accusations that he was really going to be damaging these kids. The global health-ethics world came in and was basically horrified by what he did.

Then George Church, who’s probably the leading biology researcher in the West right now, actually gave a counterargument, and he said, “No, people need to do this. We need to actually try these things.” I’m not even taking a position on this. I’m not even saying these things are good or bad, or should be allowed or should be banned, or whatever.

They’re damned interesting.

Everything I just described is becoming possible, and these things have just incredible implications for everything from health to society, which will play out for hundreds of years to come. I think more and more people are probably going to realize that there’s a lot more to discuss on those fronts than we’ve been doing.

2 closing questions. The first is around you and your firm having coined this concept of American dynamism. Increasingly, defense technology, which used to be a backwater for early-stage investors, has now become an incredibly popular category, with lots of capital flowing into it. People are thinking about national defense, an evolving war zone that has a lot more massing of drones and fewer soldiers, and all these different changes.

Is there anything in that world that you’re most attuned to personally and learning the most about today?

11. The Future Of Warfare

American dynamism covers a lot of territory, and we do a lot in there. We’re doing more in energy. We have our first nuclear investment. We’re doing education and other fields, and so there’s a lot in there, this theme of upgrading America.

For me, the most fascinating thing is the change in military affairs—the change in the nature of war and defense that’s coming from the rise of technologies like AI and autonomy. I highlight a couple of things there. One is that it’s actually happening, and in particular, the tragedy of the Ukraine battlefield also turns out to be a living laboratory for extremely rapid evolution of military technology.

The Ukrainians and the Russians have both adapted their technology enormously in the last few years to both attack and defend. Of course, there’s incredible use of drones and incredible innovation happening there. Military planners are watching what’s happening on that battlefield very closely.

The entire nature of defense systems is changing in real time in response to watching that play out.

The loose concept we have on that is that wars in the past were won by the side that had the most men and material. So if you had the biggest army and the most weapons, you won. Probably in the future, wars are won by the people with the most money and the most technology. The reason is because if you've got the most money, you can buy the most technology or develop it.

But the future of warfare probably has a lot more to do with machines fighting with each other than people. There's an obvious social welfare benefit to that, which is you'll have a lot fewer soldiers, sailors, marines, and pilots dying in war. But it changes the dynamics of the calculus of what it means to go to war. Maybe the fear would be that it makes it easier to go to war because the human cost is lower. Maybe countries become bolder and more willing to enter military conflicts more aggressively.

So there's all kinds of questions and implications there, but this shift is happening. It's going to happen. China has a massive program. A lot of what China's focused on in AI is being able to apply it in military settings. So they have a whole program on that, and then our defense establishment has the same thing, and many other countries have the same thing.

That's going to be very dramatic. I was talking to a very senior military planner a while ago, and he said, in his view, the weaponized drone that fits in a backpack, that you can fly over a hill and that can basically destroy a tank—what they call a suicide drone—was the biggest innovation in defense technology since the stirrup.

Wow.

He said the stirrup was a big deal because it was the thing that took a soldier who previously had to get off his horse to attack somebody to the horseman being able to stay on the horse, stand up, and fire a bow and arrow. The stirrup was a massive extension of individual lethality on the battlefield. And he said a squad of well-trained human soldiers with drones—20 human soldiers with drones—should be able to hold off thousands or tens of thousands of regular troops. It's a fundamental change in the economics of attack and defense. It still feels like we're on the front end of trying to process what that means.

So as we don't end there, I'll ask a quick final question. I love how you are always reading these incredible books and then taking from them frameworks to apply to your understanding of the world. If you could leave everyone with 1 book that has an interesting framework beyond The Machiavellians that you mentioned earlier, what would you pick?

I'm still very locked into this book called The WEIRDest People in the World, the Joseph Henrich book, which is probably a decade old now. But I think that book never quite gets a lot of attention. That book is extremely insightful into the nature of cultures, and in particular, the nature of different cultures. As we're in this more globalized world with all these geopolitical conflicts, it's a very insightful look into what makes cultures and, in particular, what makes Western cultures.

As you know, so much of our politics has to do now with the nature of Western cultures and what it means for immigration, all the different debates around that, and so forth. To me, it's been the most informative book to try to understand how to think about cultures.

Awesome place to close, Mark. Thank you so much for your time.

Good. Thank you, Patrick. Awesome. Thanks for having me back.