OpenAI has announced a national program it calls "Grid Protectors." Under it, the company says it will work with utilities, regulators and policymakers to design its future compute campuses so they use less electricity when the grid is under strain. The first test is a project already under contract: OpenAI's planned campus in Effingham County, Georgia, near Savannah, which Georgia Power expects to need about 3,200 megawatts.

OpenAI's announcement is dated Nov. 2, 2026, and Latitude Media analyzed it on Nov. 3. The program describes a set of intentions, not a binding contract. Latitude Media called it non-binding and said it does not break any new ground. The Effingham pilot is different, because it is tied to a signed utility agreement. Georgia Power describes its load flexibility agreement with OpenAI as "among the largest single-facility demand response commitments in the country."

Below is what the program says, how the Georgia deal is structured, and which parts are commitments and which are projections.

What is OpenAI actually committing to?

According to OpenAI's program page, Grid Protectors is meant to produce arrangements under which future OpenAI campuses reduce their demand on the grid when the system is stressed. OpenAI says it will work with utilities to agree on how much demand can be reduced, how quickly and for how long. It also lists tools it may use, depending on local needs and available equipment: supplying power from on-site generation or storage, and using batteries and power controls to smooth changes in demand. The arrangements will differ from one location to another, and OpenAI says it will share additional commitments as they are finalized.

OpenAI also makes a broader policy argument on the page. It says long-term contracts and financial assurances can protect other customers if a project is delayed, reduced or canceled. It also says flexibility cannot replace investment in new energy supply and transmission. That second point limits the claim: OpenAI is not saying flexible demand removes the need to build more supply.

Latitude Media's reading was skeptical. Its Nov. 3 analysis noted that the program is non-binding and that demand flexibility is not novel. It pointed to the AI Energy Management Alliance, formed in September by Google, Nvidia and Emerald AI to advocate for policies that promote flexibility. Latitude also quoted Drew Maloney, president and CEO of the Edison Electric Institute, a power industry trade group, on the industry's interest in large loads that can be more flexible during periods of system need.

How does the Effingham County pilot work?

The Georgia deal itself is older than the program. Georgia Power announced on July 22 that it would serve the OpenAI project in Effingham County. Latitude Media identifies the campus as Project Camellia, where OpenAI is beginning its move into developing data centers directly, a shift Latitude says the company announced in July. Latitude reports that operations are scheduled to begin in phases starting in 2028.

Georgia Power's release sets out the main terms:

  • Load: Georgia Power says OpenAI is expected to need about 3,200 MW.
  • Flexibility: OpenAI has agreed to provide up to 1,000 MW of flexible demand response, according to Georgia Power. Under the 25-year agreement, the utility can reduce the energy it delivers to the facility at certain times.
  • Costs: Georgia Power says OpenAI will pay the full cost of the infrastructure and electric service, backed by long-term energy contracts and financial assurances. OpenAI's program page makes the same commitment.
  • Regulatory framework: Georgia Power says this approach is aligned with the large-load rules the Georgia Public Service Commission first approved in January 2025, and that it reviews large-load contracts with PSC Staff.

Aaron Mitchell, Georgia Power's senior vice president of strategic growth, is quoted in the utility's release.

The flexibility figure is large compared with the campus. Up to 1,000 MW is close to a third of the roughly 3,200 MW Georgia Power expects the site to need. OpenAI compares the 1,000 MW to the average electricity demand of roughly 700,000 households. That comparison is OpenAI's own.

"Up to" matters. The figure is a ceiling on the demand response OpenAI has agreed to provide. It is not an amount of power that will be shut off on any particular day. Neither source says how often curtailment would happen or for how long.

What about the savings figure?

OpenAI's program page cites a Georgia Power projection of approximately $950 million in annual customer savings beginning in 2029. OpenAI also cites Georgia Power's estimate of a benefit of at least $15 a month for a typical residential customer using 1,000 kWh.

Two caveats apply. First, this is a projection from the utility, cited by OpenAI. It is not a measured result. Second, OpenAI says the figure comes from Georgia Power's portfolio of large-load agreements and expected growth, including Project Camellia, not from the Effingham campus alone. The sources reviewed here do not split out how much, if any, of the projected savings comes from the OpenAI contract specifically.

The sources also do not give a total dollar cost for the Effingham campus or the infrastructure needed to serve it, so DataCenterFeed is not reporting a figure.

Why does demand response matter for a project this size?

For a utility, the hardest part of serving a multi-gigawatt load is the periods of peak demand, such as heat waves or cold snaps. A load that can be reduced on request during those periods puts less strain on the system than one that has to be served at full power at all times. That is the argument OpenAI's program makes. Georgia Power says the demand response resource will let it invest in fewer new generation resources to serve future growth.

OpenAI's own page adds that flexibility does not replace new supply and transmission. A campus that can cut up to 1,000 MW at times of need still needs power the rest of the time. Under the terms Georgia Power describes, OpenAI pays the full infrastructure and electric service costs to serve its facility.

The local politics around data centers are also part of the context. Latitude Media describes the announcement as coming amid a growing public backlash over electricity costs and on the eve of midterm elections, and cites a Ballotpedia count of at least 50 local ballot measures related to data centers in 13 states. Neither OpenAI nor Georgia Power ties the program to those measures.

What is new here and what is not

The Effingham contract terms, including the 3,200 MW load, the up-to-1,000 MW of demand response, the 25-year term and the full-cost commitment, were in Georgia Power's July release. The new part is OpenAI turning that structure into a named program, saying Effingham will serve as a pilot and that future campuses will have arrangements that differ by location. Latitude Media's assessment is that the announcement breaks no new ground and is non-binding.

What Happens Next

Latitude Media reports that operations at the Effingham County campus are scheduled to begin in phases starting in 2028. Georgia Power's projected customer savings, as cited by OpenAI, begin in 2029. Georgia Power says it reviews large-load contracts like this one with Georgia PSC Staff. OpenAI says arrangements for other campuses will vary by location and that it will share additional commitments as they are finalized.

Sources