U.S. electric utilities are seeing an unprecedented surge in power requests driven by Big Tech’s scramble to build new data centers to support AI computing needs, Reuters news report said.
A Reuters survey of 13 major utilities revealed that nearly half have received inquiries for electricity volumes that surpass their entire current peak demand or generation capacity, illustrating the colossal scale of demand looming from AI data infrastructure.
Utilities are now wrestling with how to respond to these requests without destabilizing the grid or overinvesting, which could lead to higher consumer costs. Many utilities have already announced billions of dollars in new capital spending, with some doubling their five-year plans.
However, estimating actual demand is complicated. Tech companies often seek bids from multiple utilities for the same project across regions, inflating projections. For instance, Oncor Electric, a Texas utility under Sempra, reported 119 gigawatts of new connection requests — nearly four times its current peak load.
PPL Corporation has received over 50 GW in data center requests, including 9 GW in advanced development, compared to its 7.2 GW generation capacity.
Evergy, serving Kansas and Missouri, saw demand nearly double to 11 GW, just above its expected system peak for 2025.
To avoid misjudging demand, companies like Oncor and PPL only allocate spending once agreements and financial guarantees are secured.
States are beginning to intervene, with Pennsylvania considering a centralized “clearinghouse” to monitor data center power requests. Industry insiders warn of risks on both sides — underbuilding could trigger blackouts, while overbuilding could saddle consumers with costs, especially if tech firms cancel plans due to rising costs, inflation, or shifting priorities.
The cost to build a megawatt of capacity has surged to nearly $12 million as of 2024, and tariffs on materials like steel may drive it even higher.
Meanwhile, the power needs of future AI models may decline. Emerging technologies, such as DeepSeek, claim to use fewer chips and less compute power, which could significantly reduce electricity and cooling requirements.
Reflecting this, Microsoft reportedly pulled back from projects totaling 2 GW of planned electricity use across the U.S. and Europe in the last six months.
These developments highlight the uncertainty utilities face as they try to balance massive demand forecasts with the volatile, evolving landscape of AI-driven data center expansion.
GreentechLead.com News Desk