OpenAI has locked in a 3.2-gigawatt power deal through 2032 for "Project Camellia," a massive data center project in Georgia powered by Georgia Power. The agreement represents a major infrastructure bet on AI compute capacity at a scale rarely seen in the industry.
The company is attempting to address community concerns about resource consumption and job creation through two distinct incentives. OpenAI pledged $80 million for local community programs and committed $71 million in Codex credits to make AI tools accessible to Georgia students. These moves signal recognition that data center projects face mounting resistance across the US.
The 3.2-gigawatt allocation is substantial. For context, that's equivalent to the power draw of a mid-sized city. The decade-long power agreement locks in supply certainty while also committing OpenAI to long-term operations in Georgia.
Resistance to data center expansions has grown as communities weigh promised jobs against actual hiring, noise, heat discharge, and grid strain. OpenAI's dual approach addresses both economic anxiety and skill gaps. The Codex credit program targets education directly, positioning the company as investing in the next generation of developers rather than just extracting resources.
Georgia represents strategic positioning for OpenAI. The state has competitive power costs and existing infrastructure. Georgia Power's willingness to allocate 3.2 gigawatts to a single customer also reflects the utility's confidence in the project's viability and its own generation capacity.
The deal runs through 2032, a 6-year commitment that suggests OpenAI expects sustained demand for training and inference compute. Data centers of this scale typically take 2-3 years to build, meaning operational capability could begin in 2028 or 2029.
Project Camellia exemplifies a broader trend. Major AI companies require massive power infrastructure that regional utilities