Skip to content

OpenAI Grid Protectors: AI campuses to cut power at grid peaks

OpenAI's new Grid Protectors Program will let its future AI compute campuses cut electricity use when the grid is strained, starting with up to 1,000 MW in Georgia.

· · 3 min read

Share
OpenAI graphic with the words Introducing the Grid Protectors Program on an orange gradient background
Image: OpenAI
On this page

AI data centres are becoming some of the largest electricity users in the US, and their impact on power bills has turned into a political issue. On 2 November OpenAI announced the Grid Protectors Program, a plan to make its future compute campuses reduce their electricity use when the grid is under strain, for example during a heat wave or a cold snap. Details come from OpenAI's announcement.

How it is supposed to work

Under the programme, OpenAI will work with utilities, regulators and policymakers to agree, site by site, how much demand a campus can drop, how quickly and for how long. Depending on local needs, campuses could also feed the grid from on-site generation or storage and use batteries and power controls to smooth out swings in demand. OpenAI says the arrangements will differ by location and that it will publish further commitments as they are finalised.

The company's argument is about fixed costs: power lines and substations cost the same whether they are busy or not. A large customer that uses a lot of electricity when there is spare capacity, and backs off at peaks, can help pay for that infrastructure without forcing expensive upgrades for the few hours of highest demand.

Georgia is the pilot

The first site is OpenAI's planned campus in Effingham County, Georgia, which is still in development. OpenAI says it has committed to pay the full infrastructure and electric-service costs for the project. Its agreement with Georgia Power includes up to 1,000 megawatts of flexible demand response — power that can be freed up during extreme heat. OpenAI puts that at roughly the average demand of 700,000 households.

Georgia Power, according to OpenAI, projects about $950 million a year in customer savings from 2029 across its portfolio of large-load agreements and expected growth, and estimates a benefit of at least $15 a month for a typical home using 1,000 kWh. Those are the utility's projections, not guaranteed results.

Why it matters

OpenAI cites forecasts that US data centre capacity will exceed 100 gigawatts by 2030. The company stresses that flexibility "cannot replace" new generation and transmission, but it wants demand response to become a standard part of how its campuses are built and run. At home, the same idea already exists on a smaller scale: many utilities pay households to let a smart thermostat ease off at peak times, and bidirectional EV charging can feed power back when the grid needs it.

It follows OpenAI's recent push to finance its compute build-out, which we covered when it cut its funding target to $20 billion.

Share

More from Computers & AI

View all →

Read next