OpenAI Raises 2030 Compute Spending Forecast to $750 Billion, Launches Self-Built Data Center Project

OpenAI has raised its forecast for compute spending through 2030 from approximately $600 billion to $750 billion. This adjustment stems from new agreements with cloud providers and the company's push to expand its own data center infrastructure.

Fact Reconstruction

OpenAI released the updated forecast around July 23, 2026, up from the $600 billion figure earlier this year. Key facts include: the company announced a $20 billion investment on Wednesday to launch the "Camellia Project," located at the Savannah Gateway Industrial Hub in Effingham County, Georgia. This marks the first data center where OpenAI is acting as the primary designer and builder, rather than simply leasing chip capacity.

Concurrent personnel changes show that OpenAI has hired Brent Mayo as Head of Data Center Construction and Delivery. Mayo previously led the rapid procurement and deployment of the Colossus supercomputer campus at xAI, and now reports to newly promoted Chief Technology Officer for Compute Capacity, Uday Ruddarraju.

Mechanism Breakdown

The direct cause of the spending increase is the signing of new cloud agreements and the advancement of self-built projects. OpenAI has signed agreements with Oracle for a total of 6 gigawatts of data center capacity, and expanded its multi-year deal with Amazon Web Services to $138 billion over eight years, including 2 gigawatts of compute using Trainium chips. Additionally, the company plans to allocate an extra $250 billion in cloud spending to Microsoft Azure.

The self-built "Camellia Project" comes against the backdrop of setbacks in the "Stargate" data center plan. That initiative, originally envisioned as a $500 billion joint investment with SoftBank, Oracle, and Middle Eastern investors, is currently struggling to gain traction. The company has completed land acquisition and signed a contract with Georgia Power to secure 3.2 gigawatts of electricity supply between 2028 and 2032. Sachin Katti noted that the company paid a significant premium to reserve the power, ensuring the utility's confidence in expanding capacity.

Internal financial pressure is another driving factor. CEO Sam Altman has publicly mentioned a $1.4 trillion compute spending figure, after which CFO Sarah Friar privately clarified to investors that the actual forecast was around $600 billion. She expressed concern that if revenue growth fails to keep pace with commitments, the company may be unable to fulfill future compute contracts.

Industry Impact

On the competitive landscape, OpenAI's strengthening of self-built capabilities will reduce its reliance on any single cloud provider, but also increases capital intensity. Anthropic has reached a multi-billion dollar server agreement with AMD, planning to purchase up to 2 gigawatts of MI450 series chips starting in 2027, with the potential for up to $5 billion in investment from AMD. This indicates that other AI companies are also alleviating compute bottlenecks through diversified chip sourcing.

For upstream and downstream enterprises, user-side developers will face more stable compute supply, but will bear indirect cost increases. Enterprise users may see growing pricing pressure for AI services, as OpenAI's cash burn is projected to reach $25 billion this year and $57 billion next year—about $30 billion more than previous estimates.

For cloud providers, Oracle, AWS, and Azure continue to secure large-scale long-term contracts, but OpenAI's self-built projects may displace some leasing demand. Power supplier Georgia Power, meanwhile, gains a firm order for capacity expansion.

Strategic Assessment

Based on existing agreements and personnel changes, the most likely scenario is that OpenAI will gradually ramp up capacity from the "Camellia Project" by 2028, while continuing to expand partnerships with existing cloud providers. Verification signals include the actual power delivery timeline from Georgia Power, the execution pace of Mayo's team relative to cloud partners' construction schedules, and whether OpenAI's revenue reaches the previously forecast $30 billion in 2026 and over $280 billion by 2030.

This assessment is solely an analysis based on publicly available facts. The alignment between revenue and spending is the core factor.