The Exchanges

Q Does that mean that money going into agentes today will be burned?

A No. I, I, let's take an example on hallucinations. So the examples I gave you were metal, medical diagnosis and law were two areas that will be unlocked once we get rid of the hallucinations. But there are, um, startups like perplexity, That are doing just fine right now, even though there's hallucinations, because it's not high risk, and it, you know, it's for entertainment only, but, but if you click those links, you can check them, and it works kind of okay. It depends on how risky the industry is that you're in, on whether or not you can get started trying to position for the wave early and generating, um, but if you're in the right position, we were in the right position For seven years, and the wave came. So that money isn't incinerated. In fact, that recent deal we just announced is more revenue than the money we've raised. So.

AI assessment note: “No. I, I, let's take an example on hallucinations.”

Q Did you ever doubt yourself? Seven years is an incredibly long wait time.

A Well, just doubt that there was doubt, but there was never a pause. And the reason was, so even back before starting the TPU, I was concerned that AI was going to be a technology that would allow some people to have outsized control. Outsized influence. If you allow that to just happen in potentially not the best hands, it doesn't really matter how rich you are. It doesn't matter. Nothing matters. It's the most important technology. So it didn't matter how hard it got. There was no choice but to be successful. And our goal is to preserve human agency in the age of AI, right? If, if we don't do that, we have failed. And so it wouldn't matter whether there was doubt or not. And yes, there was plenty of doubt. There was a point where we were so close to running out of money. We did this thing that we called Grok Bonds. So you know, um, um, war bonds from World War II?

AI assessment note: “there was doubt, but there was never a pause”

Q Can you just walk me through how that deal is structured?

A Yeah. So we started off last year, right? And, and we got to, um, 19,000 of our chips deployed. We did that in about 51 days. And The question was, what can we do this year? So they've gone off, they've collected up a bunch of power in the country, and the deal is structured so that they will put up the capex for us to deploy our chips in that data center, or those data centers, and we pay back based on the money that we make. So it's, it's sort of a, it's a little bit different than debt, In that they participate in the upside, um, but it, it's similar in nature, but it is revenue, because we actually make profit up front.

AI assessment note: “the deal is structured so that they will put up the capex”

Q you know, tool, which allows you to prompt and middle build the app. I'm sure you've seen bolt.new. I'm not sure if you've seen lovable, where it's basically chat GPT, but for kind of website creation in its bluntest terms. Is there value in that? Everyone was like, there's no value in these wrapper apps. Everyone's like, there's no value in these foundation models. Where the fuck is that value?

A And that's part of the exciting part. It's discovering that. But I think people will always prefer to use the highest quality, most polished product. I think there is an opportunity for artisanship, craftsmanship, right? And just perfecting it. Right? And, and getting to a certain number of nines in the details. I mean, like the Eames quote, the details aren't the details, the details are the thing. I used to be a little concerned with the quote, um, you know, if you're not ashamed of the quality of your, your first release and you've waited too long, because there's a subtlety and nuance there. There's soundness and then there's completeness. What you want is an incomplete product, something that doesn't do everything. That's why you should be embarrassed, but it shouldn't like blue screen of death on you, right? That that's not a good, um, embarrassment, right? And so what you're going to see now is because it's so easy to come up with something that just kind of works. It's a little embarrassing, but it kind of works. People are really going to value well-crafted, high quality, products.

AI assessment note: “People are really going to value well-crafted, high quality, products.”

Q You said there about kind of booking, you know, um, IP addresses from China. Um, there's a lot of concern about US customer data going back to China. Do you think that is a legitimate and justified concern?

A Yes, it's, it's, um, it's probably the most significant concern. There, there are other concerns that's probably the most significant because people don't think they're so used to using these services. When, when you use one of these other services, you might be shocked to hear this. When you say delete, what they do is they write delete right next to your data. They don't actually delete it. They just market delete it. When you later come back and ask for your data, they give it to you with the word delete right next to it. It's still there. And these are well-meaning companies. Do you really think like the CCP doesn't have all your data and isn't going to look it up later? And some governments are more aggressive than others, right? And if they have access to your data, not even your data, it could be your next door neighbor's data. Your next door neighbor might put something in there That, um, accidentally, um, gives information away that makes you more vulnerable. Right. And then now the CCP has something and like, maybe you had some package delivered and, and they put a complaint somewhere and whatever, like you might not even do it yourself, but other people around you, like the health data of a spouse.

AI assessment note: “Yes, it's, it's, um, it's probably the most significant concern.”

Q Listen, dude, I could talk to you all day. I do want to do a quick fire. What do you believe that most around you disbelief?

A I'm going to go with anti-founder mode here. I'm anti-founder mode. I believe in delegation. I think when you are telling people how to do their job, that is an indication that it's not a, not necessarily a problem with you. It could just be that that person is not right for that job. And it's much easier to just direct them than to go find someone else competent. But it also means you probably haven't aligned them. So we, we align people through this challenge coin. Um, so everyone at Grok carries this, um, twenty five million token per second challenge coin. And what this is, is it tells everyone what we're doing. It's an alignment. And I can't tell you how many people I've showed that to who are like, that's awesome. And yet no one else is making them.

AI assessment note: “I'm anti-founder mode. I believe in delegation.”

Q Okay, what's up with the five hundred billion dollar Stargate effort? Do you buy, there's numbers of bullshit.

A I've gone back and forth on that. I, I actually did, so Gavin Baker tweeted some math. Before I saw that tweet, I came up with very similar math. Like, spookily similar math. However, Talking to some people in the know, some of the comments are actually, they've got it. But then you keep pressing and it's like, well, maybe is there some cutesiness to it? What I think it is, is an acknowledgement that the models have been commoditized. And infrastructure is what's important in terms of maintaining, uh, elite, like scale. It's one of the seven powers. And so I think what you're seeing there is an attempt to move from having a cornered resource or something like that into a scale economy.

AI assessment note: “What I think it is, is an acknowledgement that the models have been commoditized.”

Q You said there about kind of booking, you know, um, IP addresses from China. Um, there's a lot of concern about US customer data going back to China. Do you think that is a legitimate and justified concern?

A Yes, it's, it's, um, it's probably the most significant concern. There, there are other concerns that's probably the most significant because people don't think they're so used to using these services. When, when you use one of these other services, you might be shocked to hear this. When you say delete, what they do is they write delete right next to your data. They don't actually delete it. They just market delete it. When you later come back and ask for your data, they give it to you with the word delete right next to it. It's still there. And these are well-meaning companies. Do you really think like the CCP doesn't have all your data and isn't going to look it up later? And some governments are more aggressive than others, right? And if they have access to your data, not even your data, it could be your next door neighbor's data. Your next door neighbor might put something in there That, um, accidentally, um, gives information away that makes you more vulnerable. Right. And then now the CCP has something and like, maybe you had some package delivered and, and they put a complaint somewhere and whatever, like you might not even do it yourself, but other people around you, like the health data of a spouse.

AI assessment note: “Yes, it's, it's, um, it's probably the most significant concern.”

Q of NVIDIA, because they dropped 16% on the thesis that the increasing efficiency means that obviously we wouldn't need as much NVIDIA chips. And I thought exactly that, which is like, you'll still need the NVIDIA inference, and you'll just have much higher usage. So to me, it's the most screaming buy of the century. Do you share my optimism on NVIDIA, given what you just said in Jevon's paradox?

A So I think over the long term, The, the only thing I say is, um, you know, Warren Buffett and Charlie Munger in the short term, the market is a popularity contest in the long term. It's a weighing machine. Um, I can't tell you about the popularity contest, but in terms of the weighing machine part, like there, this is a misunderstanding. It's actually more valuable. Thanks to deep seek, not less valuable. Okay. So Jevin's paradox was actually discovered by, by Jevin and, and You know, um, as, as recently made famous in Satya's tweet. However, I did beat him to that, um, by quite a bit. And, and just as Satya likes to say that he made Google dance, I'm gonna say I made Satya dance, right? But he might take exception to that. But, you know, less than a month before he posted that, I, I did a, a cute little tweet on it. Um, so What's really happening here was in the 18 sixties, this guy Jevin, he actually wrote a trees on steam engines, which I guess is what you did for fun back then in, in England. And, um, he realized every time, uh, steam engines became more efficient, people would buy more coal, which is the paradox. But if you think about it from a business point of view, when the OpEx comes down, more activities come into the money. So people do more things, right? And, and so what's happened is every time we've seen the cost of tokens for a particular level of quality of mo…

AI assessment note: “It's actually more valuable. Thanks to deep seek, not less valuable.”

Q different guests. And then we supplement it with a huge amount of research from speaking to Chamath, And speaking to Scooter and speaking to everyone in between. We care about it being good enough first and then fucking great later with all the references. Most people will actually just be happy with good enough and get away with it. Do we as a human society get happy with good enough?

A When we hire, we hire for something that we call booking the win early. So one of the most important Driving forces for people is loss bias. When you have something, you don't want to lose it. People are less likely to go after something that they, they've already had, and you, you did grow up in a, in a family that was well off, and then you lost that. That might be part of the drive because you wanted to get back to it. When we have an engineer that we're hiring, and there's a room full of people who are saying, you know, if we do this thing, we could be twice as fast. I want that engineer to hear, wait, if we don't do that, we're going to be half the speed we could have in. The loss bias, right? Book the win early. Because it's possible, it must be done. I think that's a smaller segment of the population. Those are the people who deliver amazing things that no one else is going to do because everyone else is like, that's good enough. However, I think with AI, it's so easy to create a prototype to stand out. You're going to need to do that. And one of the things that's happened with the ability to communicate more freely and see what other people are doing, like, can you think back to what the restaurant experience was 20 years ago? Versus what it is now. The average restaurant is better than what high-end restaurants were 20 years ago. Because people see all of the, the stuf…

AI assessment note: “Those are the people who deliver amazing things... because everyone else is like, that's good enough.”

Q Does it? So, so for the chip providers today that are coming out, and we, we are seeing new chip providers come out where they're raising like a lot of money from good people. It's too late.

A Yeah, so the reason that I decided to go into chips, um, I did the Google TPU, but, um, also before I left, I set a record on the, uh, best classification model, like ResNet-Fifty, um, uh, with someone in Google Brain. We, we did an experiment. We, we beat everything, um, and so I could have gone in the algorithm side, and the reason that, um, and, and actually when we were fundraising, I wasn't even a hundred percent sure that I was gonna do chips. I was, like, thinking maybe we'd We do something on the algorithm side, especially in formal reasoning, uh, which is good that I didn't, but, um, the, the main motivation to go into chips was the, the moat, the, the temporal moat. So a question we get asked by VCs a lot is what prevents someone from copying what we're doing? And the answer to that is if you copy what we do, You're three years behind us because it takes that long to go from the design of a chip to a chip in production if you execute perfectly. I've done, um, three chips now that are, um, in production or ramping to production. All three were A zero silicon. Only 14% of chips that are taped out the first time Work the first time, our A zero silicon. So that means there's an 86% chance each time that you're gonna have to re-spin it. When we built our V two chip, we actually, um, uh, already scheduled a re-spin for it. And we ended up not having to do it because to our …

AI assessment note: “If you copy what we do, You're three years behind us”

Q You know, obviously my mother has MS and that's incredible. It was always taught to me that it was incurable. And actually like now it's like, actually, maybe not. What are you singly most excited for?

A We went from a phase where people were hardware engineers to, they were software engineers. To be a hardware engineer is ridiculously difficult. The, the training, you have to get things right. There's a real expense if you get it wrong. Becoming a software engineer So much easier. All you have to do is get a little bit of time on a machine, and you can teach yourself. Nowadays you can just download, um, manuals from the internet, um, or, or tutorials from the internet, or, or whatever. I think prompt engineering is going to unlock a huge swath of human society. There's 1.4000000000 people in Africa who know how to, who know how to speak, and if you were to give them access to a tool that They could create applications live just by speaking to it. That would be another 1.4000000000 potential entrepreneurs. There's eight billion people on the planet. And the difference is hardware was just ridiculously difficult. It was arcane knowledge that's hard to get. Software was plentiful. Language, you already know it. You don't have to learn a thing. What's that going to do for venture? What's that going to do for entrepreneurialism?

AI assessment note: “I think prompt engineering is going to unlock a huge swath of human society.”

Q To who though? Given the concentration of their buyers. Shit.

A Um, so no one is successfully predicting how fast AI, okay, we started off talking about is AI a bubble? If you look, uh, for the last 10 years, infrastructure for data centers, you're planning that out two, three, four, five years in advance, right? And what happens is everyone, everyone's predictions are wrong. They end up building too little. This has just been what, what's happened for the last 10 years. So you don't build enough for 10 years. What do you do? You try and overbuild. You try and build more than your most optimistic projections. And then once again, you haven't built enough. So you increase your projections and you just keep doing this. Um, that's what's been happening, and yet people still aren't building enough compute. And where people's, um, instincts are off, and, and this just hasn't, I think, been recognized yet. AI doesn't work the way SAS does. In SAS, you have a bunch of engineers who go out and build a product, and the quality of that product is determined based on what those engineers did. That's not the case in AI. In AI, I can improve the quality of my product by running two instances of the prompt, and then picking the better answer. I can actually spend more to make my product better on each query. I can even decide this customer's more valuable, and I'm going to give them a better result. That's kind of what OpenAI announced when they said, ah…

AI assessment note: “we started off talking about is AI a bubble?”

Q junior to mid-level in some of these companies. And they are living an amazing life, actually, in great places. You think they're living that amazing life in Guangdong when they're working for DeepSeek or any other Chinese alternative? I don't think so. I think they're actually getting paid much less working that fucking ass off 20 hours a day and not getting kombucha and being paid two million a year.

A That not only fair, we have a policy that we never offer the highest because we want people to choose us, not choose the salary. If we win in a bidding war, then that means the next time someone comes along with a higher salary, that's it. They're just going to go take that other job. There's no loyalty. They don't believe in the mission. Instead, we focus on, look, we're going to build this. This is your opportunity. You're going to get to work with amazing people. Um, Spend some time with the team. Are these the people you want to be working with? Because frankly, you're going to make so much cash. It doesn't matter. But bet on the equity, the outcome, right? Help us make this thing valuable. And people who buy into that, they're so much easier to manage because they're mission oriented. They all want to do the same thing. They're not there because they want the kombucha and they're not going to complain because the, the cappuccino machine is broken. They'll just go and buy their coffee next door.

AI assessment note: “we have a policy that we never offer the highest because we want people”

Q Do you think NVIDIA move into the model providing?

A It's possible, but I think, I mean, if I was them, I would avoid it because I wouldn't want to give the customers of mine. I mean, NVIDIA is great at training, right? It, it, it's crazy. It would be like, you know, um, being a, um, automotive, a, a, a car company and then creating your own taxi service. You're now competing directly with your customers, right? And I think tech companies love to do this. We have a management philosophy, and it's based on big O complexity, and we only do things that require a sub-linear number of employees. So what I mean by that is, if someone comes to me and says, I need 10 people to go do this thing, A lot of people would say, well, why can't you do it with five? I would say, okay, you're supporting customers. If we double the number of customers, do you need 20 or do you need 11? Because I want to know what's that growth rate. Are they automating everything, right? We completely automated our compiler. We completely automated everything that, you know, large portions of our cloud. And that means that we can scale with a small team. We have 300 people. We have 300 people. We built our own chip. We built our own networking hardware and software. We built our own runtime. We built our own orchestration layer. Um, we built our own compiler. We built our own cloud. We built all this with 300 people. Now, we would only be able to do this with a sma…

AI assessment note: “It's possible, but I think, I mean, if I was them, I would avoid it”