OpenAI has raised its planned compute spending through 2030 to roughly $750 billion, up about 25% from the approximately $600 billion figure the company had set earlier in 2026, the Wall Street Journal reported. The revision reflects both rising per-unit training and inference costs and the infrastructure required to support hundreds of millions of active users across ChatGPT, Codex and the company's growing enterprise product line.
The increase also reflects a strategic shift: OpenAI is moving from being purely a compute tenant, renting capacity from Microsoft Azure, Oracle and CoreWeave, toward building and owning data-center capacity directly. That shift was already visible in OpenAI's $30 billion Project Camellia campus in Georgia, announced days earlier, and this compute-forecast revision suggests Camellia is one piece of a much larger direct-ownership strategy rather than an isolated project.
The timing is notable: OpenAI's upward revision lands the same week Alphabet raised its own 2026 capex guidance to as much as $205 billion and saw its stock drop roughly 7% anyway, and the same week Tesla's capex-driven earnings miss helped trigger a broader market selloff. OpenAI, as a private company, doesn't face the same immediate public-market reaction -- but the revision shows the entire frontier-AI industry is still scaling its compute commitments upward even as public investors grow visibly more skeptical of exactly this kind of open-ended spending.
For infrastructure, power and semiconductor investors, a quarter-trillion-dollar increase in a single lab's multi-year compute forecast is a more durable demand signal than any individual data-center groundbreaking -- it implies sustained, multi-year demand for power procurement, specialty construction, cooling technology and chip supply regardless of which specific vendor wins any given contract.
Watch whether OpenAI discloses how much of the $750 billion will be self-funded versus externally financed, and whether other frontier labs -- Anthropic, Google DeepMind -- follow with comparable multi-year compute forecast increases of their own.
