OpenAI lifts compute spending plan to $750B
OpenAI raised its infrastructure spending plan through 2030 to roughly $750 billion, up from about $600 billion earlier this year, as new cloud deals stack up.
OpenAI has raised its projected infrastructure spending through 2030 to around $750 billion, up from roughly $600 billion earlier this year, according to Wall Street Journal reporting picked up July 22. New deals with cloud providers are named as the driver of the increase, on top of existing commitments to companies including Oracle, AWS, and Broadcom.
The revision lands the same week OpenAI said it would spend $20 billion building Project Camellia, a data center in Effingham County, Georgia, that marks the company’s first turn as principal designer and builder of its own facility rather than leasing chip capacity from a cloud partner. Reporting also says OpenAI’s own CFO, Sarah Friar, has privately flagged a risk that the company may not be able to honor its future computing contracts if revenue growth doesn’t keep pace with the spending commitments.
The number matters less as a single figure than as a signal of how tightly OpenAI’s unit economics are now tied to its own procurement pipeline. That’s the same dynamic explored in what the codex economics paper means: compute commitments made today become the pricing floor tomorrow, and a lab that has pre-committed three-quarters of a trillion dollars has less room to cut prices than one still buying capacity on the spot market. If your roadmap assumes OpenAI’s per-token pricing keeps falling at the pace it has, this is a data point that the assumption has a ceiling, not a reason to panic about availability today.