Where Is the AI Compute Profit Pool Moving?

Nikesh Arora says the AI compute profit pool lives in NVIDIA and the memory industry and is shifting toward hyperscalers that already hold large amounts of compute. Availability of compute will matter more than the chips. Demand is insatiable, with more demand than Nikesh Arora and Chamath Palihapitiya can imagine.

Eight years ago, when he came to Palo Alto, he shut down all the data centers. The company could not run a global footprint at the scale, speed, and reliability he could rent from hyperscalers. Compute is now highly scarce, so he takes terms rather than dictates them. Building his own data centers is an empty threat. Nikesh Arora says Jensen cannot give him chips for 2028, and land, power, and a shell are hard to get. He does not run 50 data centers around the world with 10 to 20 megawatts of inference.

The likely path sits in the middle. He expects bare metal from a hyperscaler or a cousin of a hyperscaler, or from somebody stitching compute from around the world to meet his needs, in the hope that he will migrate to something more resilient and more robust over time. He expects hyperscalers to offer bare metal and let customers run open source models, which they already do to some extent, because the profit pool is what moves.