01:08:47 · Dylan Patel, Jordi Hays, Max Kan
[01:08:47] S3: That makes sense. Well past the point where we're trying to measure, you know, our models similar to humans. It's like we're trying to measure when will models be superhuman because we're not gonna pay for models that are similar to, like, regular humans. People are only gonna pay for models that are superhuman in certain domains and then utilize them to advance humanity. [01:09:07] S0: I'm so curious. We were talking about this earlier on the show. How how should people read into Elon's post? I believe it was Saturday that that, like, Grok bot is just too expensive to run. Sorry. Tangent. But I'm curious what the what the house take is on that. [01:09:25] S2: Like, it might be too expensive to run today, especially when consider sort of, like, the CPU and, you know, sandboxing cost needed per user. But I think one thing to understand about these personal agents is that it's very likely you will not need sort of the smartest frontier model a year from now in order to, like, serve the majority of people. And sort of the the cost to serve will actually undergo this, you know, potential 100 x plus. I follow the same trend as essentially a decreased cost of intelligence that Dylan was talking about. So even if, like, personal agents are really sensitive to serve today, I think it's very likely that they are totally economical to serve a year
01:09:25 · Max Kan, Jordi Hays, Andrew Megalaa
[01:09:25] S2: Like, it might be too expensive to run today, especially when consider sort of, like, the CPU and, you know, sandboxing cost needed per user. But I think one thing to understand about these personal agents is that it's very likely you will not need sort of the smartest frontier model a year from now in order to, like, serve the majority of people. And sort of the the cost to serve will actually undergo this, you know, potential 100 x plus. I follow the same trend as essentially a decreased cost of intelligence that Dylan was talking about. So even if, like, personal agents are really sensitive to serve today, I think it's very likely that they are totally economical to serve a year [01:10:02] S0: from now. I know. So what's the logic? You have a product that people love that's like, you know, gaining traction and then you basically, like, have everyone hit their usage limits, you know, suddenly and then say, oh, it's too expensive. [01:10:17] S5: Like has Elon's [01:10:18] S0: been on the record saying that he thinks that Grok bot will one day, and again, I don't know how much to read into this, but he's like, he thinks that Grok bot will will consume the majority of their compute over a longer time horizon. So like, what what what... You know, assume the cost is gonna drop a 100 x, like, why do you why do you stop, like, subsidizing that when it's seemingly working?
01:10:42 · Dylan Patel, Jordi Hays
[01:10:42] S3: Well, Elon, you have to remember what Elon's trying to do right now. Right? He's taking SpaceX and turning it into a Neo Cloud. Right? In in in some sense. Right? He's building ridiculous amounts of compute. And, you know, he's only got... You know, if you talk take his balance sheet, you know, he's got, what, like, $80,000,000,000 of cash on it. $20,000,000,000 is gonna go to, you know, rockets and stuff like that. But then the other 60,000,000,000 he can spend on compute, he's raising another 40,000,000,000 right now. At the same time, he claims he wants to build 13 gigawatts. 13 gigawatts, if you build all of that out, is is, you know, you know, 400 plus billion dollars. Right? You know, 500 plus billion dollars. So so where does he get the cash? Right? He has to monetize his compute as effectively as possible to able to pay for all the CapEx that he wants to do. And given what's going on in the spot compute market, right, you know, being able to sell GPUs for, you know, $15 an hour, when the TCO of them is only, like, $4 an hour, you you have this massive delta. And so he's exploiting that, and he needs to focus on how do I build the most compute, period. Right? And that's that's step one. Step two is, okay. How do I use that compute most effectively to dominate the universe? Right? It's sort of like, I don't know if you're, like... You take a evil look at it. I'm sure it's just to... It it is just to benefit humanity. But the only way [01:11:57] S0: you can build 13 gigawatts a little bit.
01:11:57 · Jordi Hays, Dylan Patel, Max Kan, Tyler
[01:11:57] S0: you can build 13 gigawatts a little bit. [01:12:00] S3: Yeah. But the only way to build 13 gigawatts or even 10 or even seven financially is if he makes as much money as possible from the compute he has. And so if that means, hey. He ripped some compute out of the x AI's team to sell it to Anthropic and Google, he had to do that. Right? Because, ultimately, a year from now, his x AI team will have a lot more compute to deliver by doing that. And and and so he needs to be able to generate all this cash flow and and and sort of bridge the financial gap to get him to being able to spend $3.04, $500,000,000,000 on CapEx over the next year or year and a half. [01:12:34] S2: Yeah. And he's gotta convince people that the the going rate for these gigawatts is not the 12 to 15, you know, billion a gig that people are signing five year contracts for. It's the, you know, 40,000,000,000 plus a gig that he was able to sell this, like, short term contract to Anthropic for earlier this year. But if you look at people who are actually all in on the personal assistant like, you know, a meta news, I I do not see them seeing, you know, stopping the subsidization anytime soon because they know it's gonna become economical at the year from now, and all that really matters at this current point in time is we we get the adoption, you know, we get the users. [01:13:08] S0: Makes sense. Makes sense. Any other questions, Tyler? [01:13:13] S7: I think... No. This is great. [01:13:15] S0: We covered it. Yeah. What else is top of mind for you guys before we jump off?
01:13:15 · Jordi Hays, Dylan Patel
[01:13:15] S0: We covered it. Yeah. What else is top of mind for you guys before we jump off? [01:13:21] S3: I I think one other thing that's really interesting is, you know, the frontier models are actually not the best at certain tasks from open and anthropic because they're safety classifiers. Right? Google and and Meta and the Chinese models are are better at, like, cyber tasks, at biotasks, at certain alignment tasks, and that presents a really interesting opportunity that certain markets should just actually not use open ananthropic unless they get into some specific sort of tent, that open air anthropic are willing to give, which plebs like, us at least are not able to. [01:13:57] S0: On, just reminded me something I wanted to ask earlier. It feels like there's a massive incentive right now for companies to come out and say, you know, we're bringing compute on prem and we're running open models and we're getting... How much should people read into that? I think the headline yesterday, I couldn't tell if it was real, is that Goldman is sort of banning the use of Claude Code and Codex internally and other coding agents. To me that seems kind of like an insane move given that they're, you know, underwriting underwriting the IPO. And it also feels like you're you're sort of betting against frontier progress and and capability increases. But, like, how should how should people think about those kind of headlines?
01:13:57 · Jordi Hays, Andrew Megalaa, Max Kan
[01:13:57] S0: On, just reminded me something I wanted to ask earlier. It feels like there's a massive incentive right now for companies to come out and say, you know, we're bringing compute on prem and we're running open models and we're getting... How much should people read into that? I think the headline yesterday, I couldn't tell if it was real, is that Goldman is sort of banning the use of Claude Code and Codex internally and other coding agents. To me that seems kind of like an insane move given that they're, you know, underwriting underwriting the IPO. And it also feels like you're you're sort of betting against frontier progress and and capability increases. But, like, how should how should people think about those kind of headlines? [01:14:48] S5: I think this is really funny because... I was... Go ahead. Yeah. I was gonna say, like, it's really funny that they would do this because, like, even on our finance, like, benchmark, it was very clear that none of the models were even able to, like, successfully do a task of an analyst. Like, the closest one got, like, 50% of the credit or something. So, like, even though if you use open models, you're just, like, shooting yourself in the foot because you're just, like, using worse models. And, like, even the frontier models aren't there. Like, imagine how much worse, like, an open source model would be. [01:15:19] S2: Yeah. But I think I think on prem slash open source, it is a real market. Is gonna keep growing, like, pretty fast in absolute terms. [01:15:28] S5: But if