As it points to the level of ambition we need to have about how much compute capacity we're going to build out, and this will be the most expensive, complex infrastructure project the world has ever collectively taken on for all of our welfare. So we're going to need a lot of compute capacity.
The good news, it would be impossible to build this out the old-fashioned way, but we'll have AI and robots helping us do this, and we'll figure it out.
And is that why space keeps coming up nowadays? There's another space race that I keep hearing about.
I honestly think the idea with the current landscape of putting data centers in space is ridiculous. It will make sense someday. But if you just do the very rough math of launch costs relative to the cost of power we can do on Earth, to say nothing of how you're going to fix a broken GPU in space—and they do break a lot still, unfortunately—we are not there yet. There will come a time. Space is great for a lot of things. Orbital data centers are not something that's going to matter at scale this decade.
But then all this infrastructure buildout, Sam, it feels like it kind of forces organizations like yours to have a very intimate relationship with government. And is that something that's kind of changing with AI? This whole power structure that you know, now at one point big tech was a very privatized force with a very complicated relationship with government. Now it seems to be that you need to have a good relationship with government, whether it's Indian companies with the Indian government or companies like yours with the US government. Is the government a big enabler?
Yeah, I think it's going to be really important, not just for building out infrastructure, but given the level of impact this is going to have on society and the need to truly democratize this technology. Governments are going to have to be involved, and companies like ours are going to have to partner with governments.
And how is that? I mean, have you ever paused and thought about that evolution of that relationship between big tech and government? By the way, can I call you big tech or not?
Really? I don't think most—I don't feel like big tech.
Okay. I don't think we're there yet.
Okay. All right. But tech companies like yours and government, I mean, I think the tech industry started out as this extremely libertarian, you know, we don't need the government, the government doesn't need us sort of view. That has changed a lot, even before AI, like the last couple of decades, as the companies got bigger and more central to the economy and the way the whole world works. That's changed significantly. But maybe never before has it been this important, just given the scale of the infrastructure that needs to happen.
And how do you feel about that? Because one of the things we keep hearing is that this White House is very close to Silicon Valley. Silicon Valley was a very close sponsor of JD Vance. That whole relationship is a very intimate one. Is that true? Is that good? Is that bad?
I would say close in some ways and not close in others. There are some tight ties, and then also this administration has had some big criticisms of tech. Is that good or bad? I think a close cooperation between tech companies and the government is going to become increasingly important over time. It obviously won't be a perfectly smooth relationship, but the better it can be, I think, the better for all of us.
And what about government-to-government relationships? You know, what got announced today was a very interesting partnership of several countries who are thinking about AI in a very advanced way. Is it all these countries versus China? Is that what's kind of playing out, or are we all kind of in a race and somewhere China is also in that race?
I suspect it'll turn out to be way more complex than that. I think it will shift over time. I do sort of think that AI will become one of the most important political issues in the world, one of the highest-order bits of geopolitical tension and cooperation. But I don't think it'll be a fixed thing. I think as it develops and as—this has always happened, political alliances shift over time. I expect them to shift more in the future, not less, as things move faster.
But how do you think about China in all of this? I mean, we keep hearing that China's leap years ahead and China's got all this energy and China's got all this renewable energy and China's got all this robotics and all this technology. Are we hearing the right thing, and why are they leap years ahead? What have they done?
I would say very ahead in some areas and not ahead in others. I think this is—I mean, that sounds like a boring statement that's probably always true in the world. Like in terms of manufacturing physical robots, clearly ahead. A big edge on things like electric motors, magnets, stuff like that. Clearly ahead on energy buildout. So there are places where China is significantly ahead of other countries, and then there's places where I think we're ahead of them, and my guess is that's sort of always what it's been like and what it will continue to be like. It's hard to be ahead on everything or behind on everything. Maybe if you have like the only superintelligence in the world, you could do it. I think that would actually be bad for other reasons.
But is it a race, Sam? I mean, we keep framing it as a race. Is it a collaborative journey or is it a race or—
Yeah. I mean, both. That's like the hard part of this, right? There are always ways in which the global economy is interdependent and very collaborative, and then there's always ways in which it's a competition. And it's so tempting to say it's all one or the other. Reality is always messier than that. But I do think it's important to look at the net differences in power and the impact that could have. And I think it's, you know, really important that democracy leads with AI.
And when you say democracy leads with AI, you mean China should not lead with AI.
Um, it was a little bit more nuanced of a comment than that, but if you want to reduce it to that—no, because in your speech yesterday, you mentioned authoritarian a couple of times. So I'm just kind of, you know, how do we think about China? How do we unpack China? What I meant more is I don't think there should be any single superintelligence in the world. I don't think there should be any one person or any one country or any one company in charge of superintelligence, including the United States. I think that would also be bad. I think the world is at its best when power is widely distributed, when people have a lot of different ideas, and when there's enough of a balance in power that we can all—countries, individuals, companies, whatever—sort of keep each other in check and all have input. And over time, even though we'll do some dumb things along the way, people kind of get to pick the best ideas and let those rise up. So what I meant by authoritarian here is you don't literally want one AI in charge of the world, no matter who has it. I think that would be bad.
Yeah. But do you think if you compare to the big tech era of seven years ago, AI is going to fragment power more or is it going to concentrate more? Because it seems to be that big tech, the old big tech, is equally sort of staying—keeping pace.
That is at some point—in some sense, I think one of the most important questions in front of us right now. You can totally imagine a world where AI massively concentrates power, where say a single company or country is able to hold on to AI and use that to amass a gigantic amount of power and wealth in the world. You can imagine another extreme where everybody on earth has a superintelligence with no rules whatsoever and some pretty terrible things and chaos happen. You can imagine many worlds there in the middle. I personally think something way more towards the democratized version of that is good. We of course will need some regulation. We of course will need some guardrails. But I think putting this technology in the hands of lots of people is possible, and that will be a decentralization of economic power. The clearest example of this you can see is in the last six, and especially three months, the power that a one- or two- or three-person company can have. This just didn't happen a few years ago. It was impossible. In this new world of things like co-dex and general-purpose knowledge-working agents, we're going to have to get used very quickly to one-person startups that have enormous amounts of success, or two-person, three-person startups. And I think this is great.
Yeah. And how competitive is all of this from your seat? I mean, we saw a little awkward moment on stage yesterday. There was several memes about it going around.
You got to give the internet something to laugh at. It is definitely competitive, but I will say that I think—
It's also very incestuous. I mean, you guys—because a lot of people were building out stuff or working with you at one point and everybody knows everybody and everybody's like Nvidia's investing in you, you're buying Nvidia chips, and then it's just there.
It is a weirdly small world for sure. I think it's very competitive commercially, but almost all of the efforts that have a very advanced model, I think, feel the gravity of what's happening, and they're very committed to getting safety and alignment right and willing to cooperate there.
A little bit more about what happened yesterday, if you want to share. This is a good opportunity.
I don't really have that much more to add. I won't push it.
Okay. You know, there's just so much to cover, Sam, and I learned so much. I'm very grateful for you because I learned so much prepping for this conversation. I was like three days of just deep into it. So I thought I'll just kind of break it up into a game. And we can get—I can just try to cover more and more stuff rather than having a conversation about it. And we'll see what we kind of do. So it's not a rapid fire because it's just too long to be a rapid fire. And I tried to make it shorter, but I couldn't. Is there any one thing you admire about Google's catch-up in the AI race? Because 10 years ago they were out of the water.
Well, 10 years ago, Google was the only serious AI effort. Five years ago, maybe—maybe three years ago when we launched ChatGPT, they were then way out of it. But the first thing I admire about Google is Demis and the Google team started working on AI before anyone else in the modern era with a lot of conviction. And I think without their inspiration we certainly wouldn't be here. So that's one thing I admire. And then a second thing, more recently, is their relentless focus and execution and ability to really scale the model after being pretty far behind is quite impressive.
Okay. Were you surprised when Apple partnered with Google for Siri?
Okay. The one country that is by and large on the right path for regulation on AI.
I don't think any of us know the answer to that yet. A thing that I'm happy about is countries are trying different approaches, and I think this is a good thing about many sovereign powers in the world. We will get to observe over the next few years very different approaches and see what works and what doesn't. And pretty quickly, I believe the world will move towards more of what works. But I'm grateful for the experimentation and different countries trying different things.
Okay. So a short game where I'll give you some criticisms of AI and you give me a short defense of each of these criticisms. So the first—
I might not have one for some. That's fine. And there's so many questions that you can always duck or pass.
Too much concentration of power.
Fair, unless we push super hard to democratize. And I think everyone in the world needs to hold countries and companies to this threshold: you don't get to say concentration of power in the name of safety. We don't want that trade. It's got to be democratized.
Okay. The amount of natural resources that are going into these data centers, the amount of water, the amount of—
Water is totally fake. It used to be true. We used to do evaporative cooling in data centers, but now that we don't do that—you see these things on the internet where don't use ChatGPT, it's 17 gallons of water for each query or whatever. This is completely untrue. Totally insane. No connection to reality. What is fair, though, is the energy consumption—not per query, but in total, because the world is now using so much AI—is real, and we need to move towards nuclear or wind and solar very quickly.
There was a—we had Bill Gates last year and I asked him this question and he gave a very interesting statistic. At that time apparently we calculated that there were 10 iPhone worth of battery that would go for every ChatGPT query, and now it's come down to one iPhone or one and a half iPhone battery life worth of energy for every ChatGPT query. Is that correct?
There's no way it's anything close to that much. It's much less than that. Way, way, way less.
Okay. All right. But his theory was that the AIs will learn from human evolution to be more efficient in how much energy they consume.
One of the things that is always unfair in this comparison is people talk about how much energy it takes to train an AI model relative to how much it costs a human to do one inference query. But it also takes a lot of energy to train a human. It takes like 20 years of life and all of the food you eat during that time before you get smart. And not only that, it took the very widespread evolution of the hundred billion people that have ever lived and learned not to get eaten by predators and learned how to figure out science and whatever to produce you. So the fair comparison is if you ask ChatGPT a question, how much energy does it take once its model is trained to answer that question versus a human? And probably AI has already caught up on an energy efficiency basis. Measure it that way.
Wow. Okay. Fantastic. AI is making my kids dumber.
True for some kids. Look, when I hear kids talk about AI, there are definitely some kids who are like, this is great, I cheated my way through all of high school, I never did any homework, thank you. And I'm like, what's your plan for the rest of your life? And they're like, well, I assume I can still use ChatGPT to do my job. This is very bad. We absolutely have to still teach our kids to learn and to think and to be creative and to use these tools. That's not what most kids say, though. Most kids say, I can't believe what I can accomplish now. Look at this thing that I've just made. I've built these incredible new workflows. Sure, like I may use ChatGPT in the way you used Google when you were in high school to help with my homework. But now that I have this new tool, I'm doing these amazing things. I think we will need to find new ways to teach and evaluate in school to make sure every kid is brought along. But the potential of this technology, the ability to learn more, do more—I have no doubt about that. When I was in school, Google had just come out, and my middle school teacher was like, this is the worst thing ever. There's no point to teaching anymore. Why do you have to memorize the date that someone in history was born if you could just look it up on Google? And my answer was, I think it's a complete waste of time to memorize what year someone was born. I will just go look that up on Google if I ever need to know it again, which usually you don't that often. And then I watched over the next few years, teachers come to peace with this, the education system evolve. And like always happens with new tools, the potential goes up and expectations go up, and we'll have to teach people to think harder, create more. And I'm pretty sure that a kid born today, when they're 18, graduating high school, will be able to do things that no one today can. And I think that's great.
The other criticism is that it's just not democratic enough, as in the people who want to pause AI don't have a voice. And I want to take a double click on this question—is that where do you see the resistance coming from? Is it coming from the people who have experienced technology at its highest levels and then they say we're nervous, we want to pause, or do you see the resistance coming from those who have never experienced and are nervous to experience and then they feel the vulnerability?
I mean, it kind of comes from—parts of it come from everywhere. Although as more people use the technology, I think there's less of a let's totally pause it. People like it, and there's more of a what is this going to mean? Can we have more of a voice? Can this go a little more slowly?
Okay. The one thing Silicon Valley should learn from Chinese technology.
Okay. The one thing Silicon Valley can learn from Indian technology.
The one real and one imagined fear of Chinese dominance in AI.
I think the imagined fear is that the Chinese military is going to send a billion humanoid robots marching through the streets and overpower the US Army or something like that. And I think that's like an example of fighting the last war and thinking it's going to be the same. I think the real fear is a new kind of war between nations that gets fought over the internet, where you influence people and hack into a bunch of critical infrastructure, and that feels totally possible. That's a very scary reality. Super worried about that. I think increasingly so people are using fear of AI going wrong to justify a surveillance state, and I don't think that people doing that have fully thought through the downsides of the surveillance state.
Yeah. Okay. You urgently need to look something up. Your phone is discharged. You borrow my phone. The letter T on my touchpad doesn't work. So you can only key in Grok, Claude, Gemini, or DeepSeek.
I guess if I have to look something up, I pick Gemini.
For other things, I'd pick other models.
Okay. The most compelling reason you have moved from a not-for-profit to a capped profit to pretty much any revenue including ads.
The need to democratize AI and the need to stay at the forefront of research—both of those require huge amounts of capital.
And do you think you're a research-first company or a product-first company?
Research-first company. The coolest thing about AI, I think, is that the great majority of making a good product is doing good research. We make a great model, the product is already pretty good. The early version of ChatGPT was almost no product at all. It was like a text box that fed into a model and then put the output on the screen. So research to continue to push what a model is capable of is what enables everything else.
Okay. The one criticism about you that bothers you when you hear it.
Yeah. I've heard basically every possible criticism, and at some point none of them bother you because you're just like, this has gotten so insane, so detached from truth. I guess the one that makes me the most sad is that he's just doing it for the power. He doesn't really care about trying to do something that's going to be helpful to the world.
What were you thinking when you decided not to take any equity in OpenAI? What was going on in your mind at that time?
That was truly one of the dumbest things—as I say this as someone who's made a lot of dumb decisions. The conspiracy theories that that brought about, like the one I just mentioned—it was totally not worth it. We were like a nonprofit. There was a kind of quirk of it where to be on the board of a nonprofit, you had to be sort of disinterested. And I had been very fortunate to have a successful career. And I was like, I don't really care here either way. But it was a very dumb thing.
And if somebody today were to offer and figure out a way for you to get into the cap table with equity, would you take it?
I mean, at this point, I feel like so tired of the whole conversation and so trapped, I'm not sure what to do. It's like a lose-lose thing for me.
Okay. So one thing—in spite of all the differences you have with Elon Musk, the one thing you admire about him. Is there any one thing you can put—all the differences—
No, I'm going to think of something, but give me a minute. He's extremely good at physical engineering and also extremely good at getting people to perform incredibly well at their jobs.
Okay. Super. Okay. ChatGPT is not known for its opinions, it's known for giving facts and factful answers back. So I'm going to ask your opinion on the news. There've been a lot of stuff in the news the last week. So we want your views on—would you support a government-ordered ban on social media for kids under 16?
I don't know if a total ban is right. Maybe it is. I think at least heavy restrictions would be good. I have a young child. I'm more interested in what I think are the very negative impacts of infinite scroll. I would like my kid not to just be one of those iPad kids going like this all the time. So that seems very important. But yeah, probably social media for kids is something we have to be very careful with.
Okay. It seems the clouds of war are again upon us. A lot of talk about Iran. If government were to ask AI companies like yours to help figure out how to get into war, how do you respond? Is Pentagon just another client for you? You're the customer, you do what they ask, or do you draw a line somewhere?
I don't think AI systems should be used to make war-fighting decisions. I don't think they're at a level of sophistication and reliability where this is a good idea. That said, we certainly want to support the government, and there's a lot of things we can do already. Someday there will be, I think, really important applications of this to defense. But right now the models have clear limitations.
It was used for Maduro's capture, I believe. Is that true?
I just don't know. I'm sure it was used in some ways. There are things that AI can do a great job of today. Like I think using AI to analyze a huge amount of intelligence reports—probably a great use of AI. And maybe it was used some way like that.
Yeah. Your views on Claude safety chief Jan Leike's dramatic resignation? I don't know if you followed that, but it made a lot of news.
I heard something about it. Like they resigned and like a letter, it's all useless or something.
Yeah. The world's going to go to hell, I want to go write poetry. It was in that direction. And a lot of safety people are getting, I think, overwhelmed. That's the real piece that I wanted you to reflect on.
