Georgia Power’s 3.2 GW OpenAI Data Center Contract Clears Regulatory Review
Georgia Power says its contract to serve OpenAI's planned 3.2 GW data center project in Effingham County has cleared the Georgia Public Service Commission regulatory process. The agreement includes up to 1 GW of flexible demand response and requires OpenAI to cover infrastructure and electric-service costs associated with the facility.
Microgrid Media Editorial Desk
Published · 4 min read
Georgia Power said on August 26 that the OpenAI data center power contract for a planned 3.2 GW computing project in Effingham County, Georgia, has cleared the regulatory process with the Georgia Public Service Commission. The development matters because the agreement combines one of the country’s largest planned single-site electricity loads with as much as 1 GW of flexible demand that can be reduced during periods of high grid demand.
Under the agreement, OpenAI will pay the infrastructure and electric-service costs required to serve the project, while Georgia Power can reduce as much as 1,000 MW of electricity delivered to the facility during qualifying periods. According to Georgia Power’s August 26 announcement, the utility expects the broader portfolio of large-load customers and projected future growth to generate approximately $950 million per year in incremental revenue-related customer savings beginning in 2029.
The agreement covers 3.2 GW of new electricity demand
The contract was filed with Georgia regulators in July and covers approximately 3,200 MW of new demand associated with OpenAI’s planned Effingham County campus.
OpenAI separately said in its Project Camellia announcement that Georgia Power is expected to deliver the 3.2 GW in phases between 2028 and 2032.
The scale makes the project significant even within the rapidly expanding U.S. data center sector. Large AI computing campuses increasingly require power levels comparable with major industrial facilities or entire cities, forcing utilities and regulators to determine how new loads should pay for infrastructure and interact with the grid.
Microgrid Media has been tracking that challenge across U.S. power markets, including PJM’s response to rapidly rising data center power demand and utility investment associated with large-load growth in the Houston region.
One gigawatt of the load can become flexible
The most consequential grid-management provision in the OpenAI data center power contract may be the commitment for up to 1 GW of flexible demand response.
Georgia Power says the arrangement allows it to reduce energy delivered to the facility during certain high-demand periods. The utility argues that this flexibility can support system reliability while reducing the amount of additional generation that otherwise would have to be built solely to cover future peaks.
The original Georgia Power project announcement describes the load-flexibility commitment as among the largest single-facility demand-response arrangements in the country.
The 1 GW figure represents nearly one-third of the project’s stated 3.2 GW maximum demand. That does not mean the facility will routinely shed 1 GW, however. The publicly available information does not specify the frequency, duration or precise operating triggers for reductions.
Cost allocation has become one of the most contested issues surrounding large data center development as utilities plan new generation, substations and transmission infrastructure to accommodate rapidly growing loads.
Georgia Power says OpenAI will pay the full infrastructure and electric-service costs needed to serve its facility and provide financial assurances under the state’s large-load framework.
OpenAI has made the same commitment publicly, saying existing Georgia families should not subsidize the project and that project-related infrastructure and service costs will be borne by the company.
The utility says incremental revenue associated with its portfolio of large-load customers is projected to provide approximately $2.847 billion in customer benefits over 2029 through 2031. Georgia Power projects that a typical residential customer using 1,000 kWh monthly could see savings of at least $15 per month beginning in 2029.
Those figures are forecasts tied to a broader portfolio of existing and projected large-load growth; they should not be interpreted as savings produced solely by the OpenAI agreement.
The project has also drawn scrutiny in Effingham County
The regulatory milestone does not eliminate broader questions surrounding the planned campus. Independent reporting has documented debate over the project’s scale, infrastructure requirements and local impacts.
Georgia Public Broadcasting reported in July that the proposed campus covers roughly 1,400 acres and could become Georgia’s largest data center development.
The public version of the utility contract also does not disclose every commercial term. Reporting during the regulatory review noted that significant portions of the agreement were treated as confidential business information.
What happens next
Regulatory clearance moves the electricity-supply agreement forward, but it does not mean the 3.2 GW load is operating today.
OpenAI says power delivery is planned in phases from 2028 through 2032. Actual demand will therefore depend on the pace at which the campus is constructed, energized and occupied.
For the power sector, the project will provide a particularly large test of whether flexible AI computing loads can become part of grid-reliability planning rather than operating exclusively as inflexible around-the-clock demand.
The next important questions are how Georgia Power implements the 1 GW flexibility commitment in practice, what generation and network investments ultimately serve the campus, and whether projected large-load revenues translate into the customer savings the utility currently forecasts.
The Microgrid Media Editorial Desk reports on microgrids, energy storage, grid modernization, data center power demand, distributed energy resources, and energy policy. Reporting published under the Editorial Desk byline is developed from primary regulatory records, utility filings, government documents, company disclosures, and independently verifiable sources, and is reviewed under Microgrid Media’s editorial standards.
Editorial oversight: Jonas Muthoni, Editor-in-Chief
