Who pays the electricity bill for AI? Data centers spark debate over rising power costs

Companies rushing to build the massive infrastructure needed for the artificial intelligence boom are facing a growing backlash over electricity prices as households and policymakers question whether data centers are driving up electricity prices.

However, a recent report from semiconductor research firm Semi-Analysis argued that data center expansion is only part of the picture, and that market design and policy decisions play a larger role in these energy price increases than AI infrastructure growth alone.

From rural Virginia to the deserts of Arizona, communities that once welcomed high-tech investments are now pushing back against data centers amid growing concerns that data center facilities built by so-called AI hyperscalers are straining local power grids and raising costs across the rest of the world.

Since 2020, U.S. residential electricity prices are expected to rise more than 36% from 12.76 cents per kilowatt-hour to 17.44 cents per kilowatt-hour in February 2026, and reach 19.01 cents per kilowatt-hour by September 2027, according to the latest forecasts from the U.S. Energy Information Administration.

“Retail electricity prices have been rising at a faster pace than inflation since 2022 and are expected to continue rising through 2026,” the EIA said in a March 2025 report before the Iran war.

President Donald Trump recently acknowledged the industry’s problems, saying data centers “need some public relations support.”

Retail electricity prices in the United States reflect the costs of generation, transmission, and transmission, as well as other factors such as taxes and public investments to replace aging infrastructure.

Semianalysis argued that an opaque market pricing mechanism known as the base residual auction accounts for a large portion of the “runaway” energy prices in the PJM interconnection region. PJM Interconnection Region is a regional transmission operator serving 13 eastern states and based in the data centers of hyperscalers such as Google, Anthropic, and Amazon.

Under this mechanism, consumers pay their expected electricity bills two years in advance to ensure sufficient power availability during peak demand periods such as heat waves and winter storms.

Future energy prices under this mechanism are predicted prices based on expected future demand, calculated through simulations performed based on proprietary models and data. However, in all predictive models, the parameters do not necessarily reflect the real-world situation.

The semi-analysis argued that PJM’s forecasts often overestimate future demand, especially as many of the region’s planned data centers face construction and assembly delays due to chronic memory shortages.

The report contrasted PJM with another energy grid overseen by the Electric Reliability Council of Texas and said prices will remain relatively stable from 2022 onwards, despite the development of data center complexes by hyperscalers such as OpenAI, Anthropic and Google.

In the United States, regulations governing the power grid are decentralized across states and utilities, and market design often determines how additional costs are passed on to households.

In its March 2025 report, EIA also noted regional price disparities, stating that regions with high household electricity prices may see price increases that exceed the national average.

“In capacity-constrained markets like PJM, prices have increased dramatically as demand for data centers has increased, but other markets have been able to have richer and more direct cost allocations,” Meghan Ruesch, a partner at Bain & Company, told CNBC.

Ruesch added that it may not always be clear exactly what is causing consumer energy prices to rise, as unrelated investments in regional grids, such as grid strengthening and modernization, and overall inflation can also weigh on household budgets.

“Even in the absence of data center investment, we expect there to be some upward pressure on price increases,” Ruesch said.

Big technology companies are also trying to address concerns about energy use, pledging to cover electricity costs for projects and develop alternative energy sources.

In January, Microsoft outlined a five-point plan that included a commitment to cover additional electricity costs resulting from investments in data centers and communities. This was followed by a similar initiative by Anthropic in February.

Most recently, President Trump summoned executives from major AI companies to the White House to confirm their pledge to protect ratepayers so that the costs of new AI data centers are not passed on to American consumers.

Chris Howard, head of data center account management at JLL, said such efforts can be important for “drawing support from communities that may be opposed to[data center]projects,” especially when data center development involves alternative investments in local communities, such as jobs and training.

But experts question the validity of these promises, given that hyperscalers have struggled to turn a profit.

“The problem is that the industry is not making profits, which puts more pressure on the industry,” said Mark Einstein, research director at Counterpoint Research.

Hyperscalers also need to clarify their plans to deal with rising electricity prices, he added. “If they keep quiet about this, the rumors will go away.”

Tech companies are also working to meet data center needs through renewable resources.

JLL’s Howard said these alternative energy sources will become increasingly important as demand for data centers increases and concerns about energy availability grow around the world.

“The average grid connection latency in major data center markets is already four to six years, with up to 10 years in cities like Tokyo,” Howard said.

Globally, these energy shortages “could create significant opportunities for energy producers, especially when it comes to renewable energy,” he added.

But skepticism about renewable energy efforts within the current U.S. administration raises questions about how far such sustainability efforts will go domestically, Howard said.

Still, analysts say it could be in companies’ interests for AI hyperscalers to deliver on these promises.

“It would definitely be better PR,” Einstein said.

But public backlash could lead regulators to impose new rules on hyperscalers, Einstein added. “That’s not what they want.”