(Jack_the_sparow/Shutterstock)
The rise of AI is putting real pressure on the U.S. power grid. As demand for compute explodes, the data centers behind AI systems are becoming some of the country’s top energy users. They are essential to digital life now, but their nonstop energy use is starting to reshape how electricity is priced and distributed across the nation.
A Bloomberg investigation found that wholesale electricity prices have spiked dramatically in areas near major data centers. In one example, prices for a single month were 267% higher than five years ago. Among the sites with the largest increases, about 75% were located within 50 miles of a dense cluster of data centers. The impact spreads through regional grids, meaning even households and businesses far from these facilities are paying more.
Behind every AI application is a constant stream of data. These centers do not just process information—they absorb, store, and move enormous volumes around the clock. GPUs run intensive training and inference jobs nonstop, while petabyte-scale storage holds model datasets. This activity produces serious heat, and the cooling systems needed to manage it drive energy use even higher. The model is always-on and built for growth, which turns each facility into a permanent power drain on the grid.
Bloomberg’s analysis points to a massive shift. By 2035, data centers supporting AI could use close to 9% of all U.S. electricity. That number would have sounded absurd a few years ago. But it is not just about how many machines are online. What matters is the movement of data, the need for speed, and the enormous infrastructure behind it. That burden, combined with aging grid systems, is sending prices upward, even in places where no AI center exists.
(fullvector/Shutterstock)
Much of the recent growth is centered in places like Iowa, Georgia, and parts of the Southwest. These regions offer cheaper land, fewer permitting barriers, and relatively easier access to power. However, many of them were never designed for this kind of industrial electricity use. Even areas with no facilities are now feeling the strain, as power from neighboring grids gets diverted to meet the load.
According to Mark Christie, a former chair of the Federal Energy Regulatory Commission, put it plainly, “The reliability crisis is here now; it’s not off in the distance somewhere.” He is referring to the blackouts and grid instability that could happen when energy demand rises faster than utilities can build the infrastructure to support it. In many cases, utilities are being flooded with interconnection requests from new data centers, driving up projected loads at a pace that outstrips planning cycles.
This pressure is not unique to the U.S. In Japan, electricity auction prices have climbed to record highs as the government prepares for a boom in AI workloads. Malaysia recently raised data center electricity rates because of local shortages. In the U.K., Aurora Energy Research estimates that power demand from AI infrastructure could drive a 9% increase in electricity prices by 2040.
Worldwide, data centers may use over 4% of all electricity by 2035, according to BloombergNEF. If these facilities were a country, they would rank fourth in energy consumption—just behind China, the U.S., and India. In the U.S. alone, AI-related data center power use is expected to grow from about 4.5% now to nearly 9% by 2035. Some experts are calling this the biggest jump in power demand since the arrival of air conditioning in the 1960s.
(BESTWEB/Shutterstock)
David Crane, CEO of Generate Capital and former Biden energy adviser, said: “Without mitigation, the data centers sucking up all the load is going to make things really expensive for the rest of Americans.” He warned that some U.S. regions could face brownouts as early as next year if demand keeps rising without a coordinated grid response.
Meanwhile, politics is shaping how the country responds. Some leaders have promised to cut electricity prices by 50%, but household bills have only gone up. Federal policies have recently pulled back support for wind, solar, and storage. BloombergNEF now expects the yearly rollout of clean energy projects by 2035 to be 23% lower than it would have been under earlier targets.
At the same time, new investment is being funneled into coal, gas, and nuclear sources, complicating the path forward. With AI infrastructure growing fast and the grid already under stress, the cost of data is no longer just measured in dollars. It is being measured in kilowatts, too.
This article first appeared on our sister publication, BigDATAwire.