OpenAI president Greg Brockman explains how the company allocates GPUs internally: 'pain and suffering' (www.businessinsider.com)

OpenAI president Greg Brockman said the company's internal fight over GPU access is an exercise in "pain and suffering," describing how compute scarcity drives intense emotion and competition across teams. He explained that OpenAI divides its GPU pool between research and applied products (the overall split set by CEO Sam Altman and applications CEO Fidji Simo), with the chief scientist and research head allocating resources inside research. At an operational level a small internal team — including Kevin Park — constantly reassigns hardware as projects wind down, responding to last‑minute requests for more capacity. The remarks underscore a core reality of modern AI: raw GPU capacity directly determines team productivity and which ideas can be pursued. OpenAI has warned that every new tranche of GPUs is immediately consumed, and Altman has signaled some compute‑heavy features will be gated by paid tiers as the company experiments with pushing infrastructure limits. The situation also reflects a wider industry trend — firms like Meta emphasize "compute per researcher" as a competitive lever — suggesting that access to GPUs, allocation policy, and cost models will increasingly shape research agendas, product rollouts, and who wins the next wave of AI advances.