OpenAI president Greg Brockman made a direct claim at the Big Technology AI Summit: if frontier model capabilities continue improving at roughly the same pace across labs, the winner will be whoever controls the most compute. His reasoning is structural, not aspirational. Demand for compute will outpace global supply, and the lab sitting on the largest hardware stack wins by default.
The argument cuts against the idea that algorithmic breakthroughs or research talent are the durable moats. Brockman is saying those advantages compress over time. What doesn't compress is physical infrastructure. That framing matters because it explains OpenAI's aggressive push into data center deals, its $500 billion Stargate commitment with SoftBank and Oracle, and its dependence on Microsoft's Azure at scale.
The full piece is worth reading for how Brockman handles the contradiction at the center of this thesis: OpenAI still has to produce the best models to justify the compute investment, and competitors like Google and Meta are not standing still on either front. The tension between 'compute rules all' and 'but the model still has to win users' is where the real argument lives.
[READ ORIGINAL →]