Meta’s AI Data Centers Face Power Grid Challenges During Winter Storms
As artificial intelligence infrastructure continues to expand across the United States, major tech companies like Meta are grappling with significant power grid challenges, particularly during extreme weather events. The recent Winter Storm Fern highlighted the growing tension between data center operations and grid reliability, raising important questions about how AI infrastructure will be powered in the future.
The Winter Storm Fern Wake-Up Call
Winter Storm Fern swept across the United States in late January 2026, leaving over a million people without power, primarily in the Southeast. As the grid struggled to meet demand during the crisis, Energy Secretary Chris Wright took the unprecedented step of authorizing data centers and other large power consumers to activate backup generators. These facilities, which typically operate independently from the grid, were instructed to contribute their “industrial diesel generators” to help stabilize power supply—generating an estimated 35 gigawatts of power.
AI Data Centers’ Massive Energy Appetite
The timing of Storm Fern coincided with a critical moment for data center expansion. The Lawrence Berkeley National Lab estimates that the share of electricity production used by data centers in the U.S. could spike from 4.4% in 2023 to between 6.7% and 12% by 2028, driven largely by generative AI demands. For grid operators like PJM, which serves the mid-Atlantic region, this growth presents unprecedented challenges. PJM expects a peak load growth of 32 gigawatts by 2030—enough power to supply 30 million new homes, but nearly all going to new data centers.
Meta’s Nuclear Power Strategy
Recognizing the energy demands of its AI ambitions, Meta has secured a massive supply of nuclear power through agreements with TerraPower, Oklo, and Vistra to energize its new Prometheus AI data center in New Albany, Ohio. The Prometheus cluster, anticipated to come online in 2026 and capable of operating at 1 gigawatt capacity, will require enough energy to power approximately 5 million homes.
Emerging Solutions: Demand Response and Virtual Power Plants
Rather than simply drawing more power from stressed grids, some companies are exploring alternatives. In August 2025, Google announced new agreements with Indiana Michigan Power and the Tennessee Valley Authority to provide data center demand response by targeting machine learning workloads and shifting non-urgent compute tasks away from peak demand times. Several new companies have also been founded to help AI data centers shift workloads and use battery storage to temporarily reduce grid demand during power shortages.
Distributed energy generation and storage, alongside winterizing power lines and using renewables, are key approaches to maintaining grid stability during winter storms. These solutions could also benefit the communities where data centers operate.
Looking Ahead
The challenges exposed by Winter Storm Fern underscore a critical reality: as AI infrastructure expands, the relationship between data centers and power grids must evolve. Whether through nuclear power partnerships, demand response programs, or distributed energy resources, the tech industry faces mounting pressure to ensure that AI innovation doesn’t come at the cost of grid reliability or public power access during emergencies.