OpenAI's new Grid Protectors program would let utilities call on future AI campuses to reduce electricity use when the grid is strained. For households watching their power bills, the November 2 announcement offers a model for sharing infrastructure costs, with the first example in Georgia still under development.
The national program builds on an existing agreement for OpenAI's planned Effingham County campus. The company says arrangements elsewhere will depend on local grid needs, including how much demand a campus can cut, how quickly it can respond and how long reductions can last. Those commitments will be announced as they are finalized.
Georgia's agreement came first
Georgia Power announced the project in July. Its expected electricity requirement is about 3,200 megawatts. A 25-year agreement gives the utility the ability to reduce energy delivered to the facility during certain periods, with up to 1,000 megawatts of flexible demand response.
That maximum reduction equals 31.25% of the project's expected demand: 1,000 divided by 3,200. It describes the scale of the commitment, rather than the amount the campus will shed during every emergency. The agreement allows reductions at specified times; it does not promise that a third of its consumption will disappear permanently.
The utility says OpenAI will pay the full infrastructure and electric-service costs needed for the facility, with long-term contracts and financial assurances intended to protect other customers.
OpenAI describes Effingham as the pilot for the national program once the campus is operating. The November announcement does not establish that the facility is already providing this service.
What the savings estimate covers
Georgia Power announced regulatory approval of the contract in August. It projects approximately $950 million in annual customer savings from its portfolio of large-load contracts and expected growth, beginning in 2029.
Its example for a typical residential customer using 1,000 kilowatt-hours a month is at least $15 in monthly savings, or $180 annually. That increased the utility's earlier $102 annual commitment by $78.
The $950 million figure covers the broader portfolio, including the OpenAI project. It is not a standalone estimate of savings created by this one campus. Nor is the $15 example an immediate credit for every US household. Both the utility territory and the 2029 start date limit what readers can infer from it.
Georgia Power attributes the projected benefit to additional revenue from large customers. A bigger customer base can help cover grid costs, but the savings depend on the projects and financial arrangements described in its forecast.
Who pays if a project changes?
Georgia's framework predates this program. In January 2025, the state Public Service Commission approved a rule allowing special terms for new customers using more than 100 megawatts.
Those terms can cover the cost of generation, transmission and distribution upstream of a site, along with site-specific infrastructure. The rule permits longer contracts and minimum billing requirements to reduce the risk that a large customer leaves before paying for facilities built to serve it. Qualifying contracts must go to the commission for review.
These financial protections address a different question from emergency power reductions. Cutting a campus's demand can help during a peak. Contract terms determine who remains responsible for the infrastructure bill.
Reducing demand is an existing grid tool
The International Energy Agency's Electricity 2026 report describes demand response as shifting or reducing consumption in response to grid or market signals. Utilities and system operators use contractual incentives to secure that flexibility. At sufficient scale, it can reduce peak-capacity needs and defer some grid investment.
The Verge's reporting places OpenAI's announcement amid growing US opposition to data centers and their energy costs. The program's promises will therefore be tested through local agreements and actual operations.
OpenAI says some locations could also supply electricity from on-site generation or storage, depending on available equipment. It acknowledges that flexibility still requires investment in additional energy supply and transmission.
