Sachin Katti, OpenAI's Head of Industrial Compute, stated that the estimated $50 billion compute spend for the year is "directionally about right," with demand still significantly outstripping available supply, leading to immediate consumption of any new capacity brought online. OpenAI's strategy involves being the tenant/offtaker for compute, pushing financing off its balance sheet while partners like Microsoft, Amazon, Google, and Oracle finance and build the underlying assets. Physical bottlenecks such as permitting, turbine supply, transformers, substations, and skilled labor are increasingly critical constraints for scaling AI infrastructure.
OpenAI's compute demand remains insatiable, driving hyperscalers to underwrite massive infrastructure buildouts, but physical supply chain bottlenecks now dictate the pace of expansion.