Data Centers and Grid Resilience: A Path to Powering Communities and AI Growth
The recent passage of Winter Storm Fern across the United States in late January 2026 underscored the critical need for robust and resilient energy infrastructure. The storm, which brought a combination of ice, snow, and frigid temperatures, left over a million people without power, with the Southeast region bearing the brunt of the outages. In response to higher than anticipated energy demand during the storm, PJM, the nonprofit organization responsible for operating the grid serving a significant portion of the mid-Atlantic U.S., sought federal authorization to increase power generation. This request included the controversial measure of allowing large power consumers, such as data centers, to activate their backup generators, even if it resulted in increased air pollution from the combustion of relatively less clean fuels.
Energy Secretary Chris Wright granted this request, taking a further step by authorizing PJM, the Electric Reliability Council of Texas (ERCOT), and Duke Energy, a major electricity provider in the Southeast, to instruct data centers and other substantial power-consuming facilities to activate their emergency generators. The primary objective was to ensure sufficient power supply to customers during the severe weather event. Typically, these facilities operate independently and do not feed power back into the grid. However, Secretary Wright explained that their diesel-powered generators possess the capacity to generate 35 gigawatts of electricity – enough to power millions of homes.
Experts in the electricity industry, based in the Southeast, view the aftermath of Winter Storm Fern as an opportunity to explore more sustainable solutions for powering data centers while simultaneously enhancing community resilience during winter storms. Data centers are significant consumers of electricity, and their escalating power demands, particularly in support of generative artificial intelligence (AI), are already contributing to rising electricity prices in already strained grids like PJM’s. Projections indicate that data centers’ electricity consumption could surge significantly in the coming years, with the Lawrence Berkeley National Laboratory estimating a potential increase in their share of U.S. electricity production from 4.4% in 2023 to between 6.7% and 12% by 2028. PJM anticipates a peak load growth of 32 gigawatts by 2030 – a capacity equivalent to supplying 30 million new homes, a substantial portion of which is projected to be dedicated to new data centers.
The rapid expansion of data centers and the need to power them have generated considerable public concern regarding potential increases in household energy costs. Additional worries include the environmental impact of power-hungry data centers reliant on natural gas generators, including concerns about air quality, water usage, and the potential for increased climate damage. Many data centers are either currently located or planned for development in communities already facing elevated levels of pollution. To mitigate these concerns, local ordinances, regulations established by state utility commissions, and proposed federal laws have aimed to protect ratepayers from price increases and mandate that data centers contribute to the transmission and generation infrastructure required to support their operations.
Beyond the increasing strain on the power grid, many data centers have requested near-constant power connections from utility companies, often aiming for uptime of 99.999%. However, since the 1970s, utilities have promoted “demand response” programs, where large power consumers agree to reduce their electricity usage during peak demand periods in exchange for financial incentives like bill credits. These programs have proven effective in helping utilities and grid managers manage electricity demand during periods of high load, particularly during the summer and winter months. The widespread adoption of smart meters has also enabled residential customers and smaller businesses to participate in these efforts, with the potential for distributed energy resources like rooftop solar and battery storage to function as “virtual power plants.”
The specific terms of data center agreements with local governments and utilities are often not publicly available, making it difficult to assess their potential for temporary demand reduction. However, uninterrupted power supply is crucial for maintaining critical data systems, including medical records, banking systems, and airline reservation systems. The current surge in data center demand, fueled by the advancements in AI, has prompted developers to increasingly consider demand response strategies. For instance, in August 2025, Google announced new agreements with Indiana Michigan Power and the Tennessee Valley Authority to implement data center demand response by prioritizing non-urgent machine learning workloads and shifting these tasks away from periods of high grid stress. Several new companies have also emerged specifically to facilitate workload shifting and utilize on-site battery storage to temporarily reduce data centers’ reliance on the external power grid during emergencies.
Research suggests that if data centers commit to flexible power usage, an additional 100 gigawatts of grid capacity – sufficient to power approximately 70 million homes – could be added without the need for new power generation or transmission infrastructure. Furthermore, researchers have demonstrated how data centers can invest in offsite generation through virtual power plants to meet their energy needs. The deployment of solar panels with battery storage at businesses and homes can enhance grid stability and provide a more cost-effective and rapid way to increase available electricity. Virtual power plants also offer flexibility for grid operators, allowing them to leverage batteries, shift thermostat settings, or temporarily curtail appliances during peak demand. These projects can also provide benefits to the communities where they are located.
Distributed energy generation and storage, coupled with winterization efforts on power lines and the integration of renewable energy sources, represent key strategies for maintaining a reliable power supply during and after winter storms. The recent experience in places like Nashville, Tennessee, where over 230,000 customers lost power during the peak of the Fern storm, highlights the vulnerability of traditional power lines. While onsite diesel generators offer an emergency solution for large consumers like data centers to alleviate grid strain, they are not a sustainable long-term solution for winter storms. Therefore, a collaborative approach involving data centers, utilities, regulators, and grid operators, focused on incorporating offsite distributed energy resources to meet electricity demand, holds the potential to lower energy prices, reduce air pollution and climate impacts, and ensure a consistent power supply for everyone, regardless of the season. However, analysts caution that the rapid expansion of the AI industry might lead to a speculative bubble, where consumers ultimately bear the cost of grid upgrades and new power generation to meet potentially unsustainable data center demands.
Source:
