Dwarkesh Patel: evaluation

Transcript context

…Right, there’s a trade-off there. But at the same time, for a solid four months, everyone was saying to OpenAI, “We’re not going to sign deals with you.” That sounds crazy, but it was because, “you don’t have the money.” Now everyone’s saying, “OpenAI, we believed you the whole time. We can sign any deal because you’ve raised all this money.” Anthropic is constrained in that sense. There are not that many incremental buyers of compute yet, because Anthropic hit the capability tier first where their revenue is mooning. That’s interesting. Otherwise you might think having the best model is an extremely depreciating asset, because three months later you don’t have the best model. But the reason it’s important is that you can sign these deals, lock in the compute in advance, and get better prices. Maybe this is an obvious point. But at least until recently, people had made this huge point about the depreciation cycle of a GPU. The bears, the Michael Burrys or whoever, have said, “Look, people are saying four or five years for these GPUs. Maybe it’s because the technology is improving so fast, but it in fact makes sense to have two-year depreciation cycles for these GPUs,” which increases the reported amortized CapEx in a given year and makes it financially less lucrative to build all these clouds. But in fact you’re pointing out that maybe the depreciation cycle is even longer than five years. If we’re using Hoppers—especially if AI really takes off and in 2030 we’re saying, “We have to get the seven-nanometer fabs up, we have to go back and turn on the A100s again”—then the depreciation cycle is actually incredibly long. I feel like that’s an interesting financial implication of what you’re saying. There’s a few strings to pull on there. One is, what happens to depreciation of GPUs? I guess I didn’t answer your prior question, which is that I think Anthropic will be able to get to five gigawatts-ish, maybe a little bit more by the end of the year through themselves as well as their product being served through Bedrock, Vertex, or Foundry. I think they’ll be able to get to five or six gigawatts, which is way above their initial plans. OpenAI will be roughly the same, actually a little bit higher based on our numbers. But anyway, the depreciation cycle of a GPU. Michael Burry was saying it’s three years or less. That’s sort of his argument. There are two lenses to look at this. Mechanically, there’s a TCO model, total cost of ownership of a GPU, where we project pricing out for GPUs and build up the total cost of a cluster. There are a number of costs: your data center cost, your networking cost, your smart hands and people in the data center swapping stuff out. There’s your spare parts, your actual chip cost, your server cost. All these various costs get lumped together. There’s some depreciation cycles on it, certain credit costs on it. You build up to, “Hey, an H100 costs $1.40/hour to deploy at volume across five years if your depreciation is five years.” If you sign a deal at $2/hour for those five years, your gross margin is roughly 35%. It’s a little bit above that. If you sign it for $1.90, it’s 35% roughly. Then you assume at that fifth year, the GPU falls off a bus and is dead. In some cases, the argument people are making is if you didn’t sign a long-term deal, because every two years NVIDIA is tripling or quadrupling the performance while only 2X-ing or 50% increasing the price… Then the price of an H100… Sure maybe the value in the market was $2 at 35% gross margins in 2024, but in 2026, when Blackwell is in super high volume and deploying millions a year, you’re actually now worth $1/hour. And when Rubin in ‘27 is in super high volume—even though it starts shipping this year, it’s super high volume next year—doing millions of chips a year deployed into clouds, you’ve got another 3X in performance, another 50% or 2X in price, then the Hopper is only worth $0.70/hour. So the price of a GPU would continue to fall. That’s one lens. The other lens is, what is the utility you get out of the chip? If you could build infinite Rubin or infinite of the newest chip, then yes, that’s exactly what would happen. The price of a Hopper would fall at a spot or short-term contract rate as the new chips come out and the price per performance goes up. But because you are so limited on semiconductors and deployment timelines, what actually prices these chips is not the comparative thing I can buy today, but rather what is the value I can derive out of this chip today. In that sense, let’s take GPT-5.4. GPT-5.4 is both way cheaper to run than GPT-4 and has fewer active parameters.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.