A national programme with one example
OpenAI said on 2 November that it is launching the Grid Protectors Program, which it describes as a national effort to make its future compute campuses reduce electricity use when the grid is under strain. The company said it will work with utilities, regulators and policymakers on arrangements that set out how much demand a campus can shed, how quickly and for how long, and that those arrangements “will differ by location”.
Only one is named. The planned campus in Effingham County, Georgia, is, in OpenAI’s words, intended “to serve as a pilot for Grid Protectors”. The campus is still in development.
The Georgia terms were approved in August
The commitment OpenAI cites is not new. Georgia Power announced on 26 August that state regulators had approved its contract to serve OpenAI’s Effingham County project, a contract filed with the Georgia Public Service Commission in July that covers 3,200 megawatts of new demand. Under it, OpenAI pays the full cost of the infrastructure needed to serve the site and has committed up to 1,000 MW of flexible demand response.
OpenAI puts that 1,000 MW in household terms, calling it the average electricity demand of roughly 700,000 homes — more, it says, than Washington, D.C. and Boston combined. That conversion is OpenAI’s own.
Whose savings
The money figures belong to the utility, not to OpenAI. Georgia Power projects approximately $950 million a year in customer savings beginning in 2029, rising to $2.847 billion across 2029 to 2031. It attributes those savings to its whole portfolio of large-load contracts and expected growth in the state, not to the OpenAI project alone. For a typical residential customer using 1,000 kilowatt-hours a month, the utility puts the benefit at at least $15 a month, or $180 a year — up from the $102 a year it committed to in December 2025.
OpenAI repeats the $950 million and the $15 figures in its announcement and attributes both to Georgia Power.
What the company concedes
The post is unusually direct about the limits of what it is offering. “Flexibility cannot replace investment in new energy supply and transmission,” OpenAI writes. It also says long-term contracts and financial assurances “can protect other customers if a project is delayed, reduced or canceled” — an acknowledgement that a campus promised to a grid may not arrive on the schedule the grid planned around.
The argument underneath is a cost-allocation one. Grid infrastructure is a fixed cost whether demand is high or low, OpenAI says, so a large customer that uses power when the system has room and backs off when it does not helps spread those fixed costs without forcing expensive upgrades to meet peak demand. The company cites a projection that US data centre capacity will exceed 100 gigawatts by 2030.
What to watch
OpenAI says it will “share additional commitments as they are finalized”. Until it does, the programme is a stated standard rather than a set of contracts: one approved agreement, at one site that has not been built, in one state. The test is whether the next campus OpenAI announces arrives with demand-response terms attached — and whether any utility other than Georgia Power signs one.
