OpenAI lines up 3.2 GW power supply for Georgia data center

OpenAI has signed an electricity supply agreement with Georgia Power for 3.2 gigawatts to support a planned data center project in Effingham County, Georgia, according to the company. The power is scheduled to arrive in phases from 2028 through 2032, putting Project Camellia among the clearest examples of how large AI developers are turning model demand into long-term utility commitments.

OpenAI did not disclose the financial terms of the Georgia Power agreement. The company said it will pay the costs tied to the project and that local electricity bills will not rise because of it. That claim matters because data center growth has become a ratepayer and grid-planning issue in multiple U.S. markets, where local officials and residents are scrutinizing who pays for new generation, transmission and distribution capacity.

The company described Project Camellia as privately funded. It also said the site will use a closed-loop water cooling system designed to limit continuing water consumption. OpenAI did not provide detailed projections for the facility’s electricity use beyond the 3.2 GW supply figure, nor did it publish operating water-use estimates in the announcement.

Those omissions will be watched. Water and power consumption have become recurring objections to AI data center expansion, especially as frontier model companies scale compute clusters beyond the size of traditional cloud regions. OpenAI has previously released only broad figures on its consumption, according to The Decoder, while criticism of the sector’s environmental footprint has intensified.

Community funding attached to the project

OpenAI also pledged $80 million for local initiatives in schools, healthcare and housing. The company separately said it would provide up to $71 million in Codex credits for students in Georgia.

The student-credit program gives OpenAI a local workforce and developer-relations angle alongside the infrastructure buildout. The Decoder noted that Anthropic and other AI companies have used similar programs, which can function both as education support and as early distribution for company platforms.

OpenAI’s package reflects the political risk now attached to AI infrastructure projects. Across the U.S., data center proposals have faced local opposition over resource use, noise and limited permanent employment after construction. The Decoder has reported that resistance has blocked large data center projects in several states.

For OpenAI, Project Camellia would extend the company’s infrastructure footprint beyond software partnerships and cloud contracts into a more explicit local bargain: a very large power commitment, a promise that residents will not absorb the utility cost, and a community benefits package. The company has not said how many permanent jobs the Effingham County site would create, when construction would begin, or what share of the power would come from new generation versus existing or planned Georgia Power resources.

The numbers that are disclosed show the scale of the bet. A 3.2 GW commitment is closer to utility-scale industrial demand than a conventional enterprise data center requirement. For the AI market, the Georgia deal is another signal that access to electricity, permits and local tolerance is becoming as strategic as model architecture or chip supply.

This story draws on original reporting from The Decoder.