OpenAI is vowing to reduce electricity use at US data center campuses it directly develops when extreme weather or other emergencies strain the power grid, Axios reported on Monday, 2 November. Axios calls the initiative the “Grid Protectors Program”.
The program builds on a commitment at Project Camellia, OpenAI’s planned campus in Effingham County, Georgia, where, according to Axios, OpenAI has pledged to cut electricity demand by up to a gigawatt when the grid needs relief. Axios says that is equivalent to the power use of roughly 700,000 homes. By our arithmetic, 1 GW is about 31% of the 3.2 GW OpenAI has said it is contracting from Georgia Power for the site.
The key insight:OpenAI is vowing to cut electricity use at US data center campuses it directly develops when the grid is strained, Axios reports, starting from a pledge to cut up to a gigawatt at its Georgia campus, which by our arithmetic is about 31% of the 3.2 GW it has said it is contracting there.
Update, 2 Nov: what OpenAI’s own post adds.OpenAI published its post on 2 November. It says that for the Effingham County campus OpenAI has “committed to pay the full infrastructure and electric-service costs required to serve the project” and that it intends “for Effingham to serve as a pilot for Grid Protectors”. Beyond cutting demand, OpenAI says it will, depending on local needs, “supply power from on-site generation or storage, and use batteries and power controls to smooth changes in demand”. The post cites Georgia Power’s projection of approximately $950 million in annual customer savings beginning in 2029 from the utility’s whole portfolio of large-load agreements and expected growth, including Project Camellia, and at least $15 a month for a typical residential customer using 1,000 kilowatt-hours; that is Georgia Power’s portfolio figure, not a saving attributed to OpenAI alone. OpenAI adds that “Long-term contracts and financial assurances can protect other customers if a project is delayed, reduced or canceled” and that “we will share additional commitments as they are finalized”. It gives no curtailment figure for any campus other than Effingham.
Axios quotes a blog post that OpenAI said it would publish later on Monday. In it, OpenAI says its “ambition is to make this kind of flexibility a core part of how we build and operate future campuses”, according to Axios. The post adds that this means “working with utilities to agree on how much demand can be reduced, how quickly and for how long”.
Axios reports that the specific arrangements will vary by campus, with details announced as they are finalized. OpenAI is casting the announcement as a move from earlier, aspirational commitments to a more practical framework for how it will operate, Axios says.
The post also argues that data centers can benefit ratepayers, according to Axios: steady AI power demand spreads grid infrastructure costs across more power sales, while flexibility can curb demand at busy times.
Up to 1 GW of demand OpenAI pledges to cut at Project Camellia when the grid needs relief, against the 3.2 GW it has said it is contracting from Georgia Power for the site. 1 GW is about 31% of 3.2 GW, our arithmetic. Sources: Axios, 2 November 2026; OpenAI’s 22 July 2026 post as reported by FourWeekMBA.
Business Pill · Grid Cost Allocation Explained
An evergreen Business Engineer explainer on how the cost of grid upgrades is allocated between large new loads and existing customers. It explains the general idea only and does not discuss OpenAI’s program.
The Georgia Template
Project Camellia’s fact sheet says “The campus is being designed to reduce power use during high demand or grid stress”. It also says OpenAI will pay all project infrastructure and electric-service costs and that the project’s service costs will not be passed to existing ratepayers.
OpenAI’s 22 July post on the project, as we reported on 28 October, said the 3.2 gigawatts would be delivered in phases between 2028 and 2032. The fact sheet lists 1,000+ jobs expected at full buildout and $80 million in community commitments over the life of the project.
What OpenAI’s Energy Lead Said
John McCarrick, OpenAI’s head of energy policy, told Axios the initiative reflects a shift in industrythinking, improving technology and community feedback. “I was in a meeting about a year ago with one of the other large hyperscalers, and they … were defending the five nines, and I don’t think anybody’s doing that now”, he said, referring to the idea that data centers must be available 99.999% of the time.
Axios notes that policymakers are already pushing flexible data centers: a FERC demand that grid operators justify or reform procedures for connecting large power users has provisions that promote it, and a Senate permitting bill would explicitly allow states to require data centers to limit power use at times as a condition of grid service.
As we read the Axios account, the new part is the operating detail: OpenAI says it wants to agree with utilities how much demand can be reduced, how quickly and for how long, campus by campus.
As we read the Georgia figures, flexibility is now part of the pitch to host communities alongside the fact sheet’s commitments that OpenAI pays the project’s infrastructure and electric-service costs.
John McCarrick, OpenAI’s head of energy policy, to Axios
“I was in a meeting about a year ago with one of the other large hyperscalers, and they … were defending the five nines, and I don’t think anybody’s doing that now”
Three Implications
FOR UTILITIES OpenAI’s post, as quoted by Axios, says flexibility means working with utilities to agree on how much demand can be reduced, how quickly and for how long.
FOR RATEPAYERS The Project Camellia fact sheet says the project’s service costs will not be passed to existing ratepayers, and Axios reports OpenAI argues steady AI demand spreads grid infrastructure costs across more power sales.
FOR POLICYMAKERS Axios reports that a FERC demand on grid operators has provisions that promote flexible data centers, and that a Senate permitting bill would let states require data centers to limit power use at times as a condition of grid service.
The framework reads: “The structural bottleneck isn’t computing, it’s electrons.”
As we read it, OpenAI’s pledge treats that bottleneck as something to manage hour by hour: the campus agrees to give power back when the grid needs it. The framework is the owner’s argument and does not discuss OpenAI’s program.
We first read Axios’s account of OpenAI’s blog post; OpenAI published the post itself later on 2 November (see the update above). Axios gives a curtailment figure only for Effingham; it does not say how much demand other campuses could cut, how often, or for how long.
The 31% share compares Axios’s “up to a gigawatt” with the 3.2 GW total OpenAI has cited for the site; the two figures come from different statements and the campus is to be delivered in phases.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
OpenAI says it wants grid flexibility to be a core part of how it builds and runs future US campuses, Axios reports, starting from a pledge to cut up to 1 GW at its Georgia site, about 31% of the 3.2 GW it has said it is contracting there, by our arithmetic.
A note on sourcing. The program, the 1 GW Effingham pledge, the blog-post quotes and McCarrick’s remarks are as reported by Axios; OpenAI’s own post, published later on 2 November, is quoted in the update box. Site commitments come from OpenAI’s Project Camellia fact sheet; the 3.2 GW figure is from OpenAI’s 22 July post as we reported it on 28 October. The 31% share is our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
