Chamath, thanks for joining 'The Axios Show.'
It's great to see you, Dan.
So you've been in Silicon Valley for about 25 years, and now it's kind of become AI Valley. What's the feeling, what's your feeling here now compared to say when you came?
Gosh, I think I mentioned this last week, but I think it's a bit of the tale of two cities here. We've always had this cycle of boom and bust. When I first came, it was right before the first .com bust. I was a 20-year-old graduate from engineering, and I wanted to be on Broadway, so to speak. Within months of getting here, it was the top of the market. We went through the .com bust and the rebuild. Went through it again in the mobile wave. Had a little burst of it in crypto that really didn't amount to much. And now this is the third big cycle. What's old is new again. The thing that's different this time is in those other cycles we were these cute, nerdy kids in California, left alone but generally trusted and well-liked. When we ventured out, people were like, 'Oh, look at these guys,' like exotic animals at the zoo. But now there's admiration but equally frustration and distrust. Certain people are viewed as pariahs, and that's hugely problematic given what this technology is capable of, which far exceeds anything we've done before. So it's a tale of two cities.
You do find this technology, generally speaking, AI, to be more important, consequential maybe within the past cycles.
Yeah, I really don't want to be hyperbolic, but I think it's true: it's the most important economic leveler of our lifetime. It will create the most equality if allowed to get to market. It allows anyone to follow their dreams and build a life outside of society's conventions. We stand on the shoulders of giants. Think of what had to happen for AI to exist: you build the internet, then the first generation of companies, they make computers and janky software, you string them together, Andreessen builds a browser, we navigate the world, Google documents all knowledge. All that is a precursor to this moment. And now AI says, 'Hey, because of all this, you can view it as training data.' I can create a magical thing that gives you all the expertise of the great thinkers in any domain, put it in your pocket, and you can use it. Not just you, but your daughter, your mom, your sister. That's what I mean by leveling.
Why is that different? You and I have smartphones. Five years ago we'd have said something similar. We have all the world's knowledge in our pocket.
Knowledge is different than expertise and intelligence. Knowledge is being able to search for the design of a PN junction. You have no idea what that means.
You can Google it and show somebody else who also doesn't understand. But you get the information. Now, you can say, 'Design me a transistor,' and get a spec that can be taped out and you receive a functional thing. That's expertise you can shape that wasn't available before. You take knowledge and make it actionable. That's the profound breakthrough. That's why it's also troubling: how did we get to a place where half the people think this is a bad thing?
Let me first ask about something you said: this leveling. We're also seeing capital inequality from folks who own these companies and their shareholders. Is there not a concern that insane amounts of wealth get concentrated in an even smaller number of people?
If we do it right, the answer will be way better than we want. Does anybody want a world where five humans win and everyone else doesn't? No, not even those five. We all want to pursue something without anyone getting in our way, with capital requirements approaching zero. And we want to live in a society where many others are doing the same, so we all appreciate we're trying to make something useful. That's where we can go.
Do you think that's where we're actually heading right now?
It's not where we are, but it's where we can go. I had a meeting with Zuck and Adam DeAngelo at Facebook, raising a round from Microsoft before the financial crisis. We had a meeting with Bill Gates. He said something like, 'A platform is when the economic value participants generate greatly exceeds that of the platform maker.' That's not what we had. But if you take that sentence, look at Google: the ecosystem far exceeds what Google makes. That's what can happen with AI if we don't mess it up. We have the ability to put a genius-level co-founder beside every human, man, woman, child, independent of age, creed, country, orientation. Real equality is when you are as smart as the smartest, as capable as the richest, as ingenious as the most connected. That's possible today. We should get out of our own way.
When you say get out of our own way, what is one thing you think is currently in the way?
Part of why Silicon Valley works is many founders have hardscrabble backgrounds and complicated relationships with their parents. It creates deep wells of insecurity. Work fills that well, leading to obsession and incredible creations. But that insecurity also spills over into grudges, enmity. In AI, there's a complicated origin story with factions, hurt feelings, lies, misrepresentations, hirings, firings, complex cap tables. It's not clean like Larry and Sergei. So what you see is personal insecurity, anger, and frustration spilling out. People can't control their proclivities to advance themselves or hurt others in doing so.
People can't control their own proclivities to advance themselves and/or hurt others.
But you've had major technologies before where leaders asked for regulation, like crypto. But you've never heard the heads of those companies say they might destroy humanity.
Nobody in crypto said 'We're the facility for assassins and child pedophilia, please regulate us.' They didn't say the negative and then ask for regulation. With AI, they say 'We may destroy humanity,' but they're not really saying 'please regulate us.' What's odd is that this was timed with fundraising cycles. They'd talk about apocalyptic characteristics before a model release and fundraising, then flip to a savior narrative, then back to doom. It's a bad strategy because it casts a valuable tool in personal issues. I find it deeply selfish.
Do you feel any part of you thinks it might happen?
No. There's no version where we're all standing around doing nothing. There is a version where we do more in a totally different way. We've dealt with transitions before. Example: Jeffrey Katzenberg and Pixar - animators were up in arms, but it led to 10x more animators. Another: radiologists - computers can detect cancer, but it led to more jobs. When we look back, we'll see it's an acceleration of higher-value work. What we'll get wrong is thinking people will stop working entirely. That's naive.
If this revolution is bigger than the internet, maybe past isn't predicate. I think about my daughter's future career - I can't think of anything certain.
Nor should you. Why define her value as a human being so narrowly? Employment matters, but are you saying she'll be unemployed and a ward of the state?
I'm saying I can't think of a job that won't be replaced, even plumbers.
Plumber: a robot can do the physical work. But who owns the plumbing business? A robotics company? Will there be no people? People will still need food, housing, clothing, and want to do more things. A thousand years ago, you had one task; now you have 35 to allocate time. In a thousand years, it'll be 300. Context switching might become an unlock. My point is not to be reductive and say it's all over.
On the doomerism side, there's also the bio weapons and war piece. Does that worry you?
Every technology is deterministic: anything possible will be done. Every invention brings good and bad together. We shape it to be overwhelmingly positive. With nuclear, we got clean energy but also Hiroshima. Same for AI. There will be nefarious uses. The question is whether they dominate. We need checks and balances. That's a more grounded debate.
When you talk about KYC, do you think companies should know who users are?
It's escalated. In a car, no ID; on a plane, yes. For models, as they become more useful, there should be different levels of expectations. If you want to send $150 million, you show ID. If a model is highly proficient, asking people to self-identify is reasonable. It allows technology to move forward with normal conversation.
You've supported President Trump. He's light touch on AI. Would he go for KYC?
I don't know, I haven't asked him. But KYC exists in pharma, commerce, treasury. It's a reasonable check for leading-edge products.
What is it like talking AI with him?
He's really inspiring, curious, smart, with off-the-charts intuition. I had a meeting about data centers and electricity rates. He scanned a 15-page deck, listened, and went right to the one key detail. I've seen that skill in incredible survivors. He's up there with the best businesspeople I've interacted with.
What's the smartest question he's asked about AI?
We don't have that many conversations, but it's constantly pressure-testing assumptions: What if models are nerfed? What if China wins? What if silicon gets stuck? He's quite versed.
Has America created a new class of oligarchs?
No. We were hurdling towards a monocultural blob - countries, boundaries, cultures, languages, religions didn't matter. Multilateral organizations with nameless, faceless bureaucrats were creating a homogenous world. That's the real issue.
who were shaping a lot of the experiences that we all had. And I can name these for you. The WTO, the WEF, I mean all of these places as, just as a handful of examples, they were stripping apart what made us all different. And whether you agree with that or not, I disagreed with it vociferously. We are in a period where the pendulum is firmly going back into the other direction. And it's not just an American phenomenon, it's every country starting to reassert its own cultural independence. And in the case of America, our political and economic independence. And it just so happens that when you talk about it that way, I mean, I gotta be honest with you, we're the dream team. Like, when you go and you wanna, you wanna project American exceptionalism, and it's like us versus France or us versus.
so that's what you view it as. So whether it be Sam or Dario, there, you view it. Whatever issues I have with the Frontier labs, and I think they could have done it better. Do I want them to win or China to win? I want them to win.
What does that mean when you say them? When I hear about the AI race, are we gonna win China? You've said it, what does winning mean?
I think in this case, you wanna have a diverse ecosystem of tool providers that can make sure that this economic empowerment, this idea of mobility, is available in a transparent way. Meaning, China has made a very important decision, and I think a very smart one, but it has a weakness. The decision that they've made is America has these closed source models, right? So we're inventing this incredible machine, but there are three or four closed machines. And as much as you wanna say that they are making a completely transparent machine, they're not, but they are making one part of it transparent. There's a little window where you can see a little bit of the gears moving. The thing is, you don't know how the gears are made, nor will you ever. So I think it's important in the future because a lot of these things become black boxes to people. You ask it a question. I'll give you a simple example. Last few weeks I've been in the middle of raising capital, you know, raising our Series A for 80, 90, you know, that's where I'm spending my time.
It's stressful, man. I don't sleep really well. I call the couple my CEO friends out and I said, you know, I've invested in you for years, but I never asked you this. Like, what's your sleep like? They were like, sleep, what the are you talking about? We haven't slept in 15 years. And it was so funny to me. So I went to ChatGPT and I uploaded my, you know, Apple Watch stuff, and I said, you know, what's going on? And it said something. My HR-V was low, all this stuff. And it's like, okay, here's why it's happening. Could you imagine perturbing those results and sending me down some weird path to take a drug or take an action? This is why we need open models or trusted actors.
But the open models is obviously where China's going. Do you think China is-
No, they're going open weight, they're not open source. You don't know how the gears are made. You have a little window into a machine where you can see the gears turning and you're like, yeah, there's-
See more of those gears than you can see of the gears downstream.
Right, so what I'm saying is we need a vibrant ecosystem of trusted actors in America. Because there are consequences for them mucking around with the gears. We're gonna get a bunch of cheap alternatives from all over the place. We're gonna wanna know who they are. And so you want 1,000 flowers to bloom here.
Can I ask about the cheaper alternatives question, which is there's so much capital that's gone into basically three companies, right? You know, OpenAI, Anthropic, SpaceX, which is now public, but still tons of capital went into xAI. Do you feel, were those the better bets, or do you think those companies eventually are gonna get undercut, but you've already seen Microsoft, for example, partnering with DeepSeek, that was just reported. Do you think we're gonna look back, I think of when you came to the Valley at some of these companies, and go, man, that was not just a bubble. They got beat by a cheaper, smarter version.
It's an incred, it's a very important question what you're asking. And I think if you go back and look at patterns of business, there's this really important pattern of business that is very, you ignore it at your own peril, which is that when you over earn, if you're an economic actor in any gain, when you over earn, it naturally attracts people to take that margin away from you. Unless you have something that's fundamentally monopolistic and closed that only you can make, you are never allowed to earn massive rates of return over above some premium over the cost of capital. Sorry, that's a lot of words. But what that is a simp, let me say it's simpler. Unless you have a monopoly like Google or a monopoly like Facebook, you'll engender competition. Especially when we all know the recipe for making these things. These are not a hidden secret.
So I'm curious, do you feel that if not the monetary valuations, there's inflation et cetera, but do you feel that OpenAI and Anthropic who are leading this race right now, will we look back in five years and they'll still be leading the race?
