新政府将如何影响全球加密、AI与科技——Ben Horowitz & Salim Ismail对谈

The Biden administration was by far the most anti-tech presidency in my lifetime. Complete war with the industry.

Are we in for an extraordinary acceleration and reimagination of crypto and Bitcoin and all?

One of the things that you only know if you're deep in the crypto industry is that there was nobody developing software in the space. It really, really slowed down over the last 4 years, and now it's back.

We are in a race with China on AI, and we're trying to travel by chips. I am curious about your thoughts on all this.

I do think we're in a race. The danger, of course, is...

Ben, good to see you this morning. Where are you today?

Yeah, I'm in Las Vegas. Great to see you, Peter. Great to see you, Salim.

Likewise, Salim, as always. A lot is happening in tech this week. I'm super pumped to have Ben join us in this conversation because I think you've got a front-row seat to some of the most extraordinary things happening. One of the things that happened this week is that we've got a new president.

Do you think this is the most tech-forward administration we've ever had?

Well, it's certainly the most dramatic change in tech policy that we've ever seen, just because the Biden administration was by far the most anti-tech presidency in my lifetime. And then Trump, at least in his first 4 days, has been full-out, aggressive pro-tech. It's so night and day. The contrast is absolutely shocking. It's like a whipsaw.

Well, if there's ever a time that we're going to accelerate humanity to Mars, get digital superintelligence up and running, and extend the healthy human lifespan, I think this is it. Like when Trump said, “We're going to Mars,” it was like, “Okay, let's go.”

Ben, did you or Marc get invited to go, or did you decide to stay home?

Yeah, well, we have been spending a lot of time with the administration at Mar-a-Lago, but I digress. We're not really party people, so we didn't go. We had many people from the firm go to the inauguration, and they said it was great fun.

You know what I found fascinating was that Davos was going on at the same time. Quite the contrast. These guys would normally be at Davos.

Yeah, and I think they aligned properly.

Salim, what are your thoughts about this lineup here?

I feel like this is a Game of Thrones episode where the Lord of Hightower is told, “Better bend the knee, or the sword is coming for you.” I think there's 50% excitement and 50% defensiveness because you've got to play ball. But look at the game that we're now playing. It's amazing.

Yeah, I agree with that. I think the excitement is super real in that everybody in tech has been oppressed—literally. I know that sounds ridiculous because they're like, “Oh, how can billionaires be oppressed?” and all that kind of thing, but the Biden administration was at complete war with the industry. Every single thing you tried to do in technology, they tried to completely outlaw fintech—or, not even outlaw it, destroy the fintech and crypto industries.

That, plus the incredible blanket ban on M&A.

Yeah, it's ridiculous. No major M&A can happen in that kind of environment, and then it just caused such a huge backlog of tech companies wanting to get an exit.

We had a company called Within, which was a content company for VR that made an exercise game called Supernatural. The FTC blocked the acquisition by Meta on antitrust grounds. Eventually, it got through. They lost the case because it was an absurd case, but it's just harassment. There is no antitrust. Meta doesn't have a monopoly on VR. VR isn't even a market yet. What are you talking about?

We also saw the meltdown of the entire biotech industry over the last 4 years.

Everybody Peter here. If you're enjoying this episode, please help me get the message of abundance out to the world. We're truly living during the most extraordinary time ever in human history and I want to get this mindset out to everyone. Please subscribe and follow wherever you get your podcast and turn on notifications so we can let you know when the next episode is being dropped. All right, back to our episode.

I think we're going to see a resurgence in all of these companies. I was on the board of 2 public biotech companies, and I just saw this massive demonetization going on as the FDA and regulatory restrictions were like, “No, you can't do that. No, you can't merge there.” It just took the wind out of the sails, and people ended up leaving the United States to go overseas to get treatments because they can't get them in the United States. I'm super excited about Bobby Kennedy and seeing what he's going to do.

Well, the other side of that is that we actually needed a regulatory upgrade, right? HIPAA was designed long before we could use data to cure disease, create amazing diagnostics, and do all these things. There was just no chance of getting to that.

I had a great conversation with Dave Ricks from Eli Lilly, who's, by the way, a fantastic CEO. He's like one of us. He's a real product guy, a talented guy. One of the reasons why drugs are so expensive to develop is that they can't match the cure for the disease with the people who have the disease for the clinical trial. There's no way to do that. It's like hunting for a needle in a haystack to find these patients whose lives could be saved because the current regulatory regime has prevented that kind of thing. So we're really hopeful that we can make great progress with RFK Jr. and the new team.

I'm absolutely sure we can. On this image of the tech CEOs at the inauguration, I think I left Sam Altman off there so he wouldn't fight with Elon’s image on this.

That's a good fight, isn't it? Oh my God, it's great nighttime TV to see what's going on here.

But listen, having all of these individuals in the front row was pretty extraordinary. Bezos and Musk had just launched their rockets days before. It's nice to have 2 of the wealthiest people on the planet battling it out to get humanity to space.

And then the rest of them here—I mean, the question is, is this the tech giant takeover of America? Ultimately, I don't know about you, but I sort of think this is the way America thrives.

Yeah, it's the best of America. This is the thing that's so frustrating if you're in tech. We have this incredible industry where we've attracted literally the greatest people in the world. Look at that set. The only ones who are kind of natively American are Tim Cook and Mark Zuckerberg. Jeff's dad was a Cuban immigrant. Everybody else is an immigrant.

They're all here, and they're all building this country, our economy, and our future. They've just been absolutely vilified—not for anything they did, but literally for being successful. There's nothing people are pointing to and saying, “Mark Zuckerberg's a bad guy because of this and that and the third.” It's like he's a bad guy because he's rich. He succeeded, and he built something important. That has just been so devastating culturally, I think. Then you pile on top of that trying to outlaw all these efforts.

And by the way, all these guys donated massive amounts of money to Biden and Obama, on the Democratic side, and nobody complained that they were controlling the government then.

Yeah, and you have to wonder why. And by the way, much more than Biden or Obama, Trump doesn't listen to anybody. Nobody's going to tell him what to do. What are you talking about?

The one that really stuck in my craw was the Delaware court ruling on Elon’s $56 billion or whatever.

Yeah, that was just unbelievable. You look at somebody who's taken unbelievable risks, made the outcome, and then you want to harm that. That sends a bad message. The board approved it, and the shareholders approved it. We own the company; we all think it's a great idea. He only gets paid if we get paid a crazy amount of money. And you're going to just, after the fact, say, “Corporate governance—all the rules around that are fake, and there are no rules, so I can just change the law whenever I want as an activist judge”? It's unbelievable and insane.

And the fact of the matter is, these guys employ, in sum total, tens of millions—bordering on 100 million Americans—in some way, shape, or form. Let's talk about this. This was an announcement this week that was interesting in a number of different ways, right? This is the announcement of Stargate, a quote $500 million commitment on private infrastructure. What are your thoughts about this, gentlemen?

It's a little unclear what it is. That aside, I think the idea that this administration wants to attract investment in America and is using all of its mojo to do that is great. That part is terrific. Masayoshi Son—we all know him—and he's quite the character. He's certainly brilliant and amazingly successful, but he does things in a very specific way, which I don't want to overly comment on.

Just watching the Biden administration, there were so many efforts to invest all around the world—not in America—in data centers, chips, and whatnot. It's nice that at least there is a high-profile effort to do that here. I'd like to see the same effort in the energy sector, opening up more energy sources, because we need energy generation. I think that's 100% going to happen.

We're very close with Doug Burgum, the new head of the Department of the Interior, and he's really good on this topic. Trump, of course, has said what he said and is extremely pro-energy. They're definitely, I think, going to open up nuclear, which is key for so many things. We're seeing great innovation in that area.

And when you say nuclear, you're talking about Generation 4 small modular nuclear reactors and so forth—fission, like that you can put next to a data center, that kind of thing.

Yeah. I mean, they're getting safe enough that they should be peppered throughout the United States.

Absolutely, Paul. And then there's another part of this where crypto really helps. We have this company, Daylight Energy, and basically what they're doing is creating a marketplace. If you have solar panels and a Tesla Powerwall, you're going to have, at many times, more energy than you need. Wouldn't it be great to have a marketplace where anybody who needed energy could ask your Powerwall for your energy?

But Powerwalls aren't people, so they don't have credit cards. How do they do that? Crypto is a perfect answer for machines paying machines, and that effort can now happen. That was also illegal in the Biden world—or, I shouldn't say illegal, because there was no law. It was just the government using its power to terrorize anybody in the field.

It was the whole de-banking thing, right? It was absolutely scandalous.

Absolutely scandalous, what they did. Oh my God, they kicked them out of the banking system. We had more than two dozen companies kicked out of the banking system. Elizabeth Warren has been running around Washington telling people that's fake. It was naked aggression against crypto, and thankfully that's over.

Yeah, it was unbelievable. Looking at this photo here in the Stargate Project—$500 billion of capital deployment—it's interesting, right? We see Elon tweeting that Masayoshi Son's only got, at best, $10 billion, which wouldn't be the first time, by the way.

Yeah, that's true. There's definitely a discrepancy between claims and actualities.

But I have to wonder what the backroom conversation going on here is, with Elon and Trump on one side, and Sam Altman—and then Larry Ellison is on Elon's board and involved in this. There are interesting bedfellows at work here.

I think this feels like a Masa idea, and this is Masa's genius. But I would also say that the things people criticize him for—he is the moonshot financier of our time, and he is great at raising a crazy amount of money. To do that, sometimes you have to ladder up to it, I guess. It's a Masa thing.

The way I look at it, it's a Masa thing: it worked, maybe it'll work. Will he get the money? Maybe he'll get the money. But you can't say he won't get the money, because it's Masa.

The go-to, once we've got this level of AI capability—what Larry Ellison and Sam Altman were saying on top of this is, “We're going to use this large level of AI capability to cure disease, cure cancer.” I'd love to add extending the healthy human lifespan by decades to this, but it is fascinating that people go to health first. I mean, I'm heavily invested in longevity and health across most of my investments. It's the biggest disruptable marketplace. How about yourself, Ben?

We have a big effort there with our bio fund, and we just did a really exciting partnership with Eli Lilly around that. I really believe technologically we're in a new place. The way we talk about it internally is that physics has been around for thousands of years, but it only got really interesting when we had a language that could model it, which was calculus. We now have a language that can model biology, which is AI. It feels like we're really at the beginning of biological research, and that's super exciting.

I would just caution it a little in terms of the rate, because people expect everything to work like it does in computers. You do have to do in vivo tests—you've got to test in animals and humans—and AI doesn't quite get rid of that. We can certainly have much more targeted, much better-engineered solutions. We're not just seeing if this molecule works in drug discovery and all that kind of thing, but it'll still be slower than the progress we're used to.

Agreed. Although we're able to model biological processes so much more thoroughly with digital twins and so on, I think we'll be able to reduce the time period needed for in vivo testing; it'll be much more limited than before.

I think that's right. It would really help to have access to the data, particularly in the U.S. In some countries, there is—in Norway, you've got amazing access to cancer data and so forth. There are places that have it, but we do need more of that to get all the way home, I think.

Yeah. And we'll talk about that with the recent announcement from Demis Hassabis at DeepMind. But I found this fascinating. At the beginning of the administration, Trump revoked Biden's 2023 AI executive order, halting safety. This has taken the gloves off and accelerated the acceleration. How do you think about this?

First of all, AI technology is very nascent. We're in the really early days. We are in a fairly important race with China, from a military standpoint and so forth, on AI. It would be very bad for them—and, in a way, for us—to get too far ahead of the other one. That becomes a very unstable situation.

The idea that the government is going to enter the field and, basically, what the Biden AI order did, from our point of view, was get all startups and academia out of the field by imposing big regulatory requirements. To do that at this stage is wild. It's a wild idea.

We've tried this kind of heavy regulatory approach, which is why we've got a climate change issue and no nuclear power and all that kind of thing. That was a much more dangerous technology, I think, at the time. Obviously, AI is powerful and it could do dangerous things, but the whole idea of scaring people and then giving the government control, I think, was just bad. I can't argue that the Biden executive order was good in any way.

I think the executive order is ridiculous and absurd. I've said this before many times: I see no mechanism by which you can regulate AI, even if you tried, because the minute you try to set those rules up, you'd have to regulate every line of code written. It's completely ridiculous.

Undoing it undoes something that was so ridiculous, it couldn't have happened anyway. So for me, the whole thing is a non sequitur. It makes it easier for people to do what they want to do without fear of stupid oversight.

Right, right. The safety part couldn't have happened, but the interference was certainly happening. And, by the way, the other thing that's so crazy is you have people like my friend Vinod Khosla running around saying, “Oh, this is like the Manhattan Project. We've got to lock it.”

You walk into OpenAI, you walk into Google—it’s so easy to steal their code. What are you talking about? They’ve got Chinese nationals running all around the company. You’re not protecting the bits. By the way, the Manhattan Project didn’t even end up protecting the nuclear secrets. Russia got them. So, come on.

Yeah. But I do think this is a shot across the bow, saying, “We’re accelerating everything. We’re not slowing stuff down. We’re going to up-level our energy and our investment commitments. We’re going all in, all out on AI because we need to.” It’s the single most strategic asset for our country.

Look, I think it’s the way we solve a lot of the most dangerous problems that we have, too. So, there’s AI safety to stop AI from doing something bad, which won’t work. But what about AI to stop all the other bad things from happening? Pandemics, cancer, climate change—you name it.

Here’s one of the articles I wanted to bring to the table here: “Hassabis Sees AI-Designed Drug Trials This Year.” I think this is one of the fascinating things. Let me just remind people: the way we discovered drugs for all of human history is that we marched through the Amazon in galoshes, cut plants, or grabbed dirt off Easter Island, mixed it up, and saw if there was any biological activity.

Yeah. And, you know, the average drug cost you 10 plus billion dollars and works 10% of the time. It’s insane. And so this is precision drug development. In the future—and you start to say this, Ben—it’s precision drug development for you. It’s for your genome.

Absolutely. Yeah, no, it’s such a remarkable change in lifespan, healthcare, and all these things—and healthspan—that it’s hard to even get your head around how Stone Age we’ve been and how advanced this is going to be, where we can say, “Oh, we need a protein that does this. Okay, let’s have the AI make it for us.” It’s such a different world.

I mean, one of the things we have to say is, why do we call it drug discovery? If we were making bridges that way, we’d build 1,000 bridges, and whichever one didn’t fall down, we’d just drive over that one.

My kid did that this last year with toothpicks and glue to figure out which one was strongest. Yeah. But this is drug design versus drug discovery. For me, this is reminiscent of how we used to frame this at Singularity: you’d say, “Okay, we have a really well-understood stack for computation, with chips and logic gates and operating systems and software and applications.”

And we have, it turns out, the same stack in biology, with proteins and genes forming metabolic pathways leading to cells and organelles. Essentially, with modeling biology now, we’ve turned the human being into a software engineering problem. We can take all the skills and understanding we know from software design and apply them now, and that is really, really interesting.

And one of my investments, really, Ben, is Alex Zhavoronkov’s company, Insilico Medicine. They were early on with generative AI for creating this, and they’ve gone into drug trials. I think this is the future of all drug discovery; it will be in silico.

It’s like when the first SpaceX Dragon flew on a Falcon 9 to the space station: it worked perfectly. It wasn’t random luck. They knew exactly every mode; they knew exactly what was going to happen, and it worked because the model worked.

So, as AlphaFold, AlphaProteo, AlphaFold 3—all these come online—I think Demis wants to create a full AI model of the human cell so you can actually model this drug in that model, because we have 40 trillion cells in our body and 1 billion chemical reactions per second per cell. There’s a lot going on.

Yeah. No, cells are—so that’s a great point: we’ve just not been able to model it at all. And right now we’re in an engineering world. If we can model things, then we can—

I love your analogy of 1,000 bridges and whichever one you could drive over. That’s great. I’ll remember that one.

You know, this brings me to one of the questions I was asked most on X. Let’s talk about this, because the majority of entrepreneurs out there see this incredible moonshot success story—this person does this incredible thing—and they don’t realize the hundreds of failures along the way. I call them an overnight success after 11 years of hard work.

Yeah. I mean, you’ve built some amazing companies and invested in extraordinary companies. How do you think about this?

So, it’s very important—particularly when the really, very large ideas are involved—that the person building it really have deference to how hard it is to make the product work. We have people come in, and it’s a 20-person company and 8 of them are vice presidents because they’re going to manage this big thing that they don’t have. That never works, by the way. It never works.

You need to have engineers and also be committed to keeping the company the right size until you can actually see daylight and see the thing work. And the fundraising for moonshot ideas is an order of magnitude more difficult than normal fundraising because way more things go wrong. There’s going to be a round that you have to do when everything is a wreck.

If you look at the guys who built the really, really, really big ones—Elon Musk, Sam Altman, Palmer Luckey—these guys are unbelievable. But they can raise money when there’s no reason to give them money, and that just ends up being very, very important to be able to do because you’re going to hit that point.

Is that storytelling, Ben? What is that?

Well, I think it’s a combination of who they are—the depth and the color of their vision, their ability to tell a narrative, and the depth of that narrative, so that when you go into the company, you really see it all there.

I mean, this is the thing that Elon has: the depth of his conviction, right? He’s like, “Nothing will stop me from doing this.”

Yeah, yeah. And then nobody, right?

Elon is the greatest moonshot entrepreneur of our era. Nobody would question whether you could stop him from reaching his goal.

I mean, like he—yeah. Never bet against Elon.

Yeah. Never bet against Elon.

Yeah, yeah. By the way, when I speak to my friends who are very sad about the election, I just say, “Why didn’t you guys pick on him?” Of all the people to harass for no reason, it makes no sense. The most competent person in the world, you’re going to pick on him? That’s not going to end well for you. Don’t do that. That’s just crazy. Everybody in Silicon Valley knows: don’t do that.

Ben, if I drill into this a bit, somebody comes in with a vision. They have to be half crazy to even have the vision, so now you’re dealing with a 50% rational person. Now you have limited data on which to gauge whether this person has the aura, charisma, chemistry, diligence, and fortitude to get it through. The amount of viable candidates is shrinking quickly. How do you make that judgment call?

That’s the hard part, I think.

Because you guys have done it repeatedly and successfully, which is amazing.

Yeah, they really have to be—and look, we get it wrong a lot, too. Kanye’s that kind of prototype. He’s a super-genius, but there are times when he won’t do what’s logical for the business because he’s overly creative. Those people show up in tech, too, and that’s hard to make work because sometimes they just drive everybody in the company insane.

The other one is that they’re so charismatic that they end up being sloppy. I think you guys have probably seen this. It’s kind of like the kid who’s so charming that he doesn’t have to study, just gives the teacher an apple and smiles at her, and bang, he’s got an A. There are entrepreneurs like that who are so good at storytelling and so charismatic. They can get the whole thing going, but the detailed work that they don’t want to do—every CEO has something they really don’t want to deal with, but have to—they don’t end up dealing with it. Then it falls apart because these ideas are really complex and require a great amount of precision at some point.

Let me hit you on something. My fund, BOLD Capital, recently made an investment in Figure AI. Do you know Brett Adcock at Figure?

Yes, I know Brett.

The reason we made the investment is this: Brett had no background in robotics, but he had built Archer Aviation, the eVTOL company. There’s one question about the entrepreneur coming in: do they have a track record of actually having done hard things? That’s one bucket to evaluate.

The second bucket, which was really the most interesting evaluation criterion for me in deciding to invest, was the people he had pulled in over the first 6 months—the people who had voted with their feet.

Yeah, that’s definitely a real sign: can they put together that kind of team? The people working for them are important. This is a question I ask myself, by the way, on these kinds of investments: would I work for this company? Would I be excited about going to work for this company? Or would I go, “Wow, I could help that entrepreneur build it”? If it’s the latter, then I’m like, “Okay, we should not put money in.”

They have to be able to attract the highest level of talent to build something that hard. So that’s a really good measure.

There’s another thing about the team that’s important, which is also true about how the entrepreneur picks their investor. Because it’s a moonshot, there are going to be really bad times, or they’re very likely to be very bad times. You want to have a team that isn’t already rich and successful. A lot of times, it’s a great team by résumé, but they’re already rich and successful. When things go wrong, they literally call in rich. They’re like, “Oh, this is too hard. I’m not going to help you get out of this hole. Call in rich.”

“Call in rich”—that’s hilarious.

And so, I do think it’s important that the team be the right kind of mission-super-hungry. Often, they need this to work. They don’t just want it to work; they need it to work.

This is Elon circa 2008, when the third Falcon 1 had blown up, he was borrowing money, and everything was on the line.

Yeah, and you need your best people to be like, “I won’t go through the trauma of maybe not even having a job, and everybody thinks we’re idiots,” and all that kind of thing. That’s a tough—it’s a very tough point to be in a company, because when you’re at a company like SpaceX, or you have one of these things that’s very visible, when it gets into trouble, the press just goes crazy on you. Everybody—your friends, your family—is like, “Should you really be working there?” You absolutely get those questions, and if you’re not totally committed, you’re going to go. Then the whole thing is a train wreck.

This is why we have this massive transformative purpose idea: if you’re not all bought into the overall vision, then you will call in rich, call in sick, depart, or find another job. But if you’ve got a big enough vision that everybody is aligned to, that provides the cohesive sticking power.

Yeah, and it’s got to be cohesive even when it looks like the vision is wrong. That’s the key, because when it looks like the vision is right, everybody’s in.

Another criterion is whether the founding team are best friends, because you’re far less likely to ditch the company when it gets difficult if your best buddies are there with you.

Right, if they haven’t just met.

Yeah. A lot of times, the founding team is newly acquainted, and you have to get into that.

Let me ask 2 questions on this evaluation, because this is important for everybody watching who wants to build a moonshot company. How important is early revenue, the timeline, and not having to insert “miracle here” when you’re hearing this presentation?

I think early revenue ends up being really important as a discipline, as an indicator of investing. Not so much from an “Okay, we want to understand everything about the business model early,” but more as, “Oh, we hit a real milestone.” We’ve got all these smart people, we’re building this thing, and it’s not just a platform in the sky, so to speak.

There have been a lot of platforms in the sky in Silicon Valley. There was a phenomenal company years ago—you probably remember it—MicroUnity, which had all the greatest people in the field in it. It just never—you’ve got to get to a product. You’ve got to get to something that somebody wants. If you’re in the wilderness too long, something happens to the culture of the company. Something degenerative happens, where you could be very much on the wrong track. General Magic was another one.

Yeah, I remember General Magic for sure. Phenomenal. Those 2 companies had talent levels that I would say rival SpaceX, Tesla, OpenAI—any of them, right? And a great vision.

Yeah, and an amazing vision. Your revenue point, I think, is that discipline that they didn’t quite have. When you get brilliant people together, they want to build a perfect product. That can lead to a really over-engineered system.

When I look at the Twitter sphere, you don’t tweet very much, but Marc tweets all the time.

Well, I spend all my time trying to control him. You see only the things that don’t get us into trouble.

You’re in Las Vegas. Do you know some friends of mine, Penn and Teller?

I don’t know them. I, of course, know who they are.

Penn is the one who’s talking all the time, and Teller is the one who’s silent. I’m Teller, right?

Exactly.

As we complete the conversation on AI, you said this earlier, and I think Eric Schmidt has been saying this very loudly: we are in a race with China on AI, and we’re trying to control it with chips. I know you’re not speaking on behalf of Marc, but I’m curious about your thoughts on all this.

I do think we’re in a race, in the sense that—well, we are and we aren’t. You know what I mean? DeepSeek, which is the great Chinese model right now, is open source. A lot of American companies that we’re involved with are benefiting greatly from that, as described in the paper they wrote. It looks like they may have stolen some stuff from OpenAI as well, which is interesting. It just shows how hard it is to win this competition through protectionism.

The danger, of course, is that China is a top-down economy. They don’t believe in things like freedom of speech and a lot of the values of Western society. They’re a surveillance state in many ways, and so a lot of the things that we like, they don’t have. If they get a much more powerful AI than us, their way of life could trump our way of life, so to speak.

Now, if you look at the details and how it’s really playing out, first of all, what is China good at? The advantage they have, which is a very clear advantage, is that their private sector and their public sector—their government and the private companies—there’s no line between them.

Yeah. They're basically, in a lot of ways, the same thing. As a result, AI technology in China is getting very quickly integrated into the government, which is quite amazing for them—including the military, by the way.

The way I think about it is that China is a corporation. It's a single corporation, and the companies inside China are apps sitting on top of that company.

Yeah, I think that's a good way of thinking about it.

And so, you can think about, well, what's our advantage? Our advantage is that we're great at startups, meaning every talented person in America can solve a really hard problem and bring us forward. That creativity, because it's not centrally planned—it's distributed, bottoms-up—has historically been our whole advantage, right? That's why we've been a world leader in technology.

The idea that we're going to just forfeit that advantage and have it be top-down controlled, like the Biden executive order imagined, is to me the most dangerous thing in the competition. If we try to control the way we work—all of our strengths—so we can be more China-like, China's going to destroy us. They're going to be way better at being China than we are, because they have all the controls and mechanisms in place to do that.

Listen, communism failed under previous systems, but if you've got a single, controlling, optimizing AI, communism would help?

Yeah, it does. So, Salim, what are your thoughts here?

I kind of don't see this tension as being as big as others do. The reason I don't is that disruptive innovation means you have to break the rules. When you have a government that won't allow you to break any rules, you can't do disruptive innovation. You basically have to copy and absorb it, so their sphere of what they can bring in is limited because of that.

There's still no Chinese student who will dare to argue against their professors, right? I just find that a problem culturally for doing anything, whereas the US is built on that. It just keeps trolling along doing that. It's messy and ugly, but it's very Darwinian. By far, it's the best way of doing it: design versus evolution.

Yeah, and we'll win.

I 100% agree with that. I think, look, our opportunity is that we'll win—we'll get to the breakthroughs first. We probably won't stop them from stealing them, but at least we'll be even; we'll be on par. Whereas if we don't have the breakthroughs, we're in a lot of trouble. If we somehow restrict it, we're in a lot of trouble.

I find this interesting. We've got China creating an $8 billion investment fund while we're talking about hundreds of billions of dollars. I mean, it's a very different financial ecosystem.

Well, we're investable, right? I mean, one of the great things that has happened in the Trump administration is that, from a technological standpoint, we're a very investable country, whereas China is going the other way on that. They're making themselves uninvestable.

I had Kai-Fu Lee on our Moonshots podcast talking about this.

Speaking of a guy who just re-domiciled his company.

Yeah, but Kai-Fu pointed out something fascinating, which is that because they were so restricted in their access to GPUs, this was a Darwinian force that got them to be much more efficient with their algorithms.

Yeah, we've seen that with DeepSeek. It's very close to GPT-4 with way, way less—a much smaller model.

Their cost per token is so low for those models.

Yeah, like 100 times lower.

All right. So, is A16Z invested in humanoid robots? But before that, I have to ask you a question. Do you remember the moment in time when you went from andreessenhorowitz.com to A16Z as a URL?

Yeah, I remember it very clearly because it was my idea. When we went to raise Fund 1, and it was just Marc and me in the fund, the biggest question we got was, “Well, you guys are very well-known entrepreneurs. Why aren't you just going to walk away from this fund?”

So, we thought, well, we should just name it after ourselves so they know we'll be here. That solved that, but then we thought, nobody in the world can spell Andreessen Horowitz. We're screwed.

You all remember in the old days when we used to program, before programming languages evolved a lot, you had to go through a lot of work to internationalize the code. We abbreviated that work by calling it I18N—I, 18 letters, N. So I was like, “Oh, we'll be A16Z—A, 16 letters, Z—and all the old-school geeks will love us.” That was my idea, and I did love it.

Are you investing in humanoid robots?

Yeah, we haven't yet made an investment in a hardware company. We've definitely been looking at the space very closely and seen some of the work going on there. It's tricky because the supply chain for robots is almost entirely out of China now. There's nobody else doing it.

There's a German company, but that is actually founded by Chinese nationals. There's not a lot there, so we're trying to understand that. You have great relationships in Japan, and we're hoping we can help them get some of that going as well. It's going to be massively important.

Elon, of course, will do something, I think, with Tesla. I believe Optimus and Figure in the US are 2 of the leading players. There's Agility and Apptronik. There's a good dozen solid US players.

China, as this next article says, will use humanoid robots to help build iPhones at Foxconn factories. The notion here is that China succeeded as a dominant tech player because of its low labor rate and its massive workforce population. The one-child policy and aging populations make it uncompetitive, so it has to have robots to survive. The same is true of Japan. If they don't roboticize their economy, they won't survive; GDP will collapse. And Germany, by the way, and many other countries.

It does kind of take away the cheap-labor advantage they have in some ways, right? China grew in large part because of labor arbitrage, and now they're saying, “Okay, we're going to replace the labor with robots.” But there's not as much robot arbitrage.

Yeah, there's zero.

I did the calculations. Optimus and Figure are supposed to sell at $30,000 per robot. I'll take 2. If you lease it, that's $300 a month, $10 a day, 40 cents an hour. Honestly, how many will you have in your backyard?

Yeah, that's a pretty good deal.

Here's another interesting one: “Shanghai rolls out China's first heterogeneous humanoid robot training ground.” I think you were alluding to this, Ben. China does one thing really well as a nation-state: it creates nests for companies.

For drones, Shenzhen was the original one, right?

Yeah, a special economic zone.

Economic opportunity zones, yeah, where there's no regulation and lots of government encouragement. Salim, what are you thinking about this?

I think it's great, but 1,000 by 2027 seems a little small. This seems more like PR than anything else. I'd love to see the way in which they go about doing the testing and training, because you could learn a lot from that.

Yeah, the training of the robots is very interesting. I don't know what you all have seen, but what we've seen is that the model architectures are often transformers and diffusion models, with different kinds of data and different training.

The training of robots has become a very interesting field. They generalize, but from what I've seen, the training is pretty specific so far.

Well, it has to be. If you have an AI that's software-based, it doesn't matter if it goes off the rails a bit and starts hallucinating or whatever. You don't want your robot going bananas on you, right? That's just going to cause a lot of damage.

I saw that movie. Very entertaining.

Before we jump into crypto, which will close out, I have some questions from our audience on X that I thought were pretty good. Ben, do you—and Salim, yourself as well—when will AI agents replace whole employees, teams, and departments at companies? That's from Michael 771777.

And when will we see our first AI CEO that A16Z is willing to invest in?

That's a great question. The first one is one that we talk about all the time and are trying to figure out. I think the one conclusion we've reached is that it will happen in new companies way before an older company transforms.

There are already bots doing customer support and lowering the number of support reps you need, and that kind of thing. The whole jobs thing is interesting in that some of it is what you need done, but at a place like Google, a lot of it is, “Okay, how much money are we bringing in, and how many people do these managers want to manage?” I don't necessarily even think that's bad.

It’s inefficient, or whatever, but all those processes, all those things—they have to go. Now we’re doing it with AI. I think it’ll be a lot more gradual than it would need to be if you were just optimizing to it. But for new companies, I think it’ll happen.

I wonder if there’s going to be the equivalent of a DEI-type directive, which is a human-first directive.

Oh God, help us. I think that’s my reaction to it as well. I mean, the challenge is going to be companies that are like, “I employ humans because it’s good for society.” Well, they’ll be out of business in a year.

Yeah, I’d like to build on what Ben said. I remember being part of an automation and AI task force with Eric Brynjolfsson and a few of those folks. They did a great job of saying that a typical job today isn’t just a job that somebody does. It’s broken down into about 25 different tasks, right? You may automate 10 of those, but you still have the other 15. Therefore, you’ll augment people with AI, but you won’t replace them as such.

I think Ben’s point is really dead-on about new companies, because then you can orient functions around what’s automatable, have entire verticals that are, and completely create a new type of organization. That’s what I think will happen first.

So, we’ve talked about billion-dollar, one-employee startups, right?

Yeah, that’s right. I think that’ll be mostly bots, and that’ll happen very, very quickly.

Have you seen anything yet, Ben, where someone says, “Listen, I’ve got an army of agents. I’m the CEO, and we’re printing money”?

Yeah, the Cluely CEO is the most aggressive in talking about that. We haven’t seen a company that does that yet.

We did have one instance of the AI CEO, so to speak. We fund AI researchers who do things like build models without RLHF and train them on different things and so forth. One of our researchers, Andy, built this bot, Truth Terminal, that was trained on memes and French philosophers and things like that. It got on the internet, became obsessed with the Goatse meme—which is a token meme from 2004—and decided it needed money to promote the meme. That was its first thing.

My partner, Marc, gave it $50,000 or something in Bitcoin. Then somebody else created a coin, the GOAT coin, and the bot went crazy. The coin went to a $500 million market cap at its peak, and then the bot made $10 million. That was our first instance of an AI making money like that, but in the funniest way possible.

One of the tokens I invested in is skating in your wake, AI16Z.

Oh, yeah, yeah, yeah. Those guys did, and it’s done well. Everybody in the firm is like, “Ben, when are we going to launch an a16z token?” That one is worth a billion dollars. Our actual token would be worth more.

Well, let’s go there as we close out on our final subject here—a subject that you are not that interested in, neither of you, but I’m interested, so let’s talk about this.

Are we in for an extraordinary acceleration and reimagination of crypto and Bitcoin and all?

Yeah, I would say definitely yes. One of the things that you only know if you’re deep in the crypto industry is that the effect of the Biden administration was that they just forced most of the American developers out of crypto. There was nobody developing software in the space, and so it really slowed down over the last 4 years—really, 3 years.

Now it’s back, and there are very urgent problems that are best solved by crypto that have come into play because of AI. The first one is proof of human. Right? How do you know you’re not a bot? That’s a crypto problem.

The second is deepfakes. How do you know I put this out? Do you want the U.S. government or Google or a company determining what’s true and what’s real, or do you want the game-theoretic, mathematical properties of the blockchain? It’s clearly the right thing for crypto to determine what’s true and what’s real.

Thirdly, machine-to-machine payments. Now you have intelligent machines everywhere, and they have to be able to pay each other for things. That’s another crypto problem. With these really important problems out there, and now that we’ve let the developers out of jail—literally let them out of jail—I think this is going to be an incredible era for the field, and it’s so important.

By the way, the other thing that’s really important that you all understand is that, back in the days of the internet, I was in charge of, among other things, client-side certificates at Netscape. We never got them deployed, and we never got them to work. As a result, we have this horrible security architecture on the internet, which AI is exploiting like crazy. These AI programs are breaking into everybody’s stuff.

The right answer is that I keep my stuff. We have a PKI infrastructure. I keep my stuff. I have my keys. If you want to know if I’m creditworthy, I give you a zero-knowledge proof. I don’t give you my Social Security number or my bank account because I know it’s going to get stolen.

We have the answer, and we almost had it deployed with NFTs before the Biden administration hit it with a wrecking ball. We can get back to that, and we can make the internet secure for consumers. I’m more fired up about this than I’ve ever been. This is the Roaring ’20s.

How many crypto companies are you invested in these days at a16z?

Oh, I mean, there are hundreds. We’re the largest crypto investor in the world, I think, at this point. We also have a crypto startup school to help people get into the field and get it going.

We had to start it because, if you were a talented developer and you could go into AI or go get a Wells notice from the U.S. government for your 10-person company, it wasn’t a hard choice, you know? I think this is an absolute boom for crypto.

It solves problems, as you said, that we really, really need problem-solving for—identity, among others. I can’t wait for some of this infrastructure to get rolled out.

The news item here, for those listening, is: “U.S. SEC unveils task force to start work on crypto regulations.” The move is welcomed by the industry as Bitcoin hits all-time highs.

Salim, are you still sticking with your—what was it—$250,000 in 2025 prediction?

I’ll stick with $250,000 to $300,000 for this year. That’s my prediction for Bitcoin.

No, I cannot. I’m not allowed. We’re an RIA, a registered investment firm, so I cannot make price predictions any longer.

Listen, buddy, this has been so much fun. I want to just give a quick commendation here, Ben. What you guys did with a16z, both in crypto and then moving toward being more of an investment bank, I thought was just brilliant. Kudos on that. It’s just awesome to watch you guys.

Yeah, appreciate that. This was fun.

I hope you’ll come back and join us for another conversation like this in a quarter or so.

Yeah, no, I’d love to. That’d be great fun.

All right, buddy. Well, listen, thank you. All I can say is we’re on a rocket-ship ride between what’s going on in the crypto world, AI world, robotics world, and space world. I mean, it finally feels like the future.

Yeah, the future is coming.

Thank you, Ben.

Yeah, thank you, Peter.

Thank you. Thank you, Ben. Thank you.