OpenAI introduced the Grid Protectors Program, which applies load shedding, a technique used on power grids since the 1970s, to data centers to cut AI energy demand during peak hours.

Electricity consumption is one of the biggest concerns residents raise when a data center is proposed in their community. To address this, OpenAI is taking steps to minimize its impact on the local electricity grid. According to the company, its Grid Protectors Program aims to reduce the power consumption of its data centers as local demand rises, so the grid is not overwhelmed during peak hours. Beyond keeping the lights on for everyone, the program would allow its data centers to connect to the grid and go online sooner, rather than waiting for additional power sources to come online.

"Whether it's heating homes during a northern cold snap or running air conditioning through a southern heat wave, families need reliable electricity. Our goal is to make campus flexibility a resource that utilities can call on when communities need it most," OpenAI said in a blog post. It added, "Through Grid Protectors, our ambition is to make this kind of flexibility a core part of how we build and operate future campuses. That means designing facilities to reduce electricity use when needed and working with utilities to agree on how much demand can be reduced, how quickly and for how long. The arrangements will differ by location, and we will share additional commitments as they are finalized."

OpenAI is testing the program at its planned Effingham County campus in Georgia. The company says it has already signed a 3.2-gigawatt agreement with Georgia Power that includes 1,000 megawatts of flexible demand response. This ensures that if demand jumps on the hottest days of the year, there will be enough power to accommodate the hundreds of thousands of air conditioners keeping Georgians cool. With the average American home drawing around 1.2 kilowatts, the 1,000-megawatt buffer is enough to power about 800,000 households, nearly 20% of the state's homes.

The technology, known as demand response, has been used for decades. Large operators reduce their demand when other consumers push the grid toward a peak. Before automation, grid operators manually called industrial customers, such as steel mills, and asked them to cut consumption to protect the reliability of the entire grid. Modern technology and IoT now make this possible automatically, allowing data centers to reduce their load in under a minute. Nvidia has already experimented with a similar system in the UK and has shown that it works.