What Must AI Labs Pay Per Megawatt in 2028?
Dylan Patel says the labs will keep buying a larger share of the world's compute, but they cannot do it at today's prices or anywhere close.
Anyone can already make money at 10 to 15 million dollars per megawatt. A GB300 rack, downloaded weights, and a listing on Open Router can bring in more than the compute costs, so that price has started to rise. To reach about 100 gigawatts combined in 2028 and take about 70 percent of the world's compute, the labs have to pay 25, 30, or 50 million dollars a megawatt. He calls that 100 gigawatt goal very aggressive.
Rules the labs argue for are slowing them more than they slow open-source Chinese models. OpenAI has not released Astra and stopped training for two weeks. Anthropic has not released the system a safety assessment called Model 2, widely believed to be the next Mythos. If revenue per megawatt stalls, the ability to outbid everyone else fades, and the labs maybe cannot reach 100 gigawatts. Where safety does not matter, they could generate 100 million dollars per megawatt or more and pay 50 million a megawatt.
Dylan Patel – Anthropic & OpenAI will have most of the world’s compute by 2028 ·
From the transcript
But ultimately, in 2028, if they get to 100 gigawatts combined, they have done
really disruptive things to the market. Because anyone can make money off of $10 to $15 million per megawatt compute today. You literally, like, I kid you not, it's not that hard. Go get a GB300 rack. Go download the Kimi weights. Go download VLM or SGLang. Set it up. You know, Codex and Fable can actually help you do this. It's pretty simple.
I mean, it's not like it's It's not trivial, but it's not like rocket science. And go put it on open router. It's very simple. And you'll start generating more revenue than you're paying for the compute. And so this has sort of already led to this compute pricing $10 to $15 million per megawatt start to inflect up. And to get to that 100 gigawatts in 2028, you have to believe...