Are data centers really driving up your power bill?

Welcome to the data center zeitgeist. Every day, a different trending news story emerges about data centers in Greenland, a growing backlash from residents in the UK and the US, or why Oracle is feeling the pressure of its $18 billion debt.

We are only just beginning to realize the true costs of viral memes like everyone creating AI '80s photos. But the political argument over AI has finally reached the electricity meter.

On September 16th, the US House of Representatives voted 417-3 for legislation that would ask state utility regulators to consider requiring data centers to cover the full cost of the new power and transmission infrastructure needed to serve them.

The measure preserves state control and amounts to a federal recommendation rather than a binding national pricing system. Even so, such an overwhelming vote shows that an issue once discussed mainly by grid planners has become a household concern.

Data centers are adding pressure, but they are not the whole story

The fear is easy to understand. Families are watching electricity bills rise while technology companies announce campuses that can consume as much power as a city. Nearly two-thirds of Americans are extremely or very concerned about the effect of data centers on energy prices, according to an Associated Press and NORC poll.

Fifty-seven percent expressed the same level of concern about water supplies. But public anxiety and causal evidence are different things. The honest answer is less satisfying than the slogans offered by either side.

Data centers are not the main cause of rising US electricity prices. In regions where demand is concentrated and spare capacity is disappearing, they are already adding billions of dollars to the cost of keeping the grid reliable. Both statements can be true.

According to the US Energy Information Administration, residential electricity prices rose from 12.52 cents per kilowatt-hour in 2014 to 16.48 cents in 2024. But the increase accelerated after 2020, when fuel costs, inflation, extreme weather, transmission investment, and utility spending increased.

Data center electricity use grew much faster. Lawrence Berkeley National Laboratory estimated that US facilities consumed about 176 terawatt-hours in 2023, equal to 4.4% of national electricity use. Its 2025 update, published in June 2026, places the 2030 range between 521 and 843 terawatt hours, or 9.5% to 15.3% of US electricity. Its reference case is 649 terawatt hours.

The figures highlight a serious planning problem, but they do not necessarily show that data centers are solely responsible for rising household energy prices. Household bills reflect a combination of fuel, transmission, distribution, taxes, and regulatory decisions. They also reflect investments approved years earlier. A data center built in one state cannot explain a fuel-price spike or a distribution upgrade in another.

In the 2026 EPRI-associated working paper, Have Data Centers Raised Your Electric Bill?, the authors estimate that data center growth modestly reduced average US retail rates between 2015 and 2024. Large customers can spread the fixed costs of power plants and networks across more electricity sales.

The price problem becomes clearer locally

National averages hide the pressure building in particular markets. PJM, the grid operator covering 13 states and Washington, offers the clearest example because it buys generating capacity in advance to ensure enough power will be available during periods of peak demand.

Its capacity price fell to $28.92 per megawatt-day for the 2024 to 2025 delivery year, then rose to $269.92 for 2025 to 2026 and $329.17 the following year. The latest auction cleared at its regulatory cap while procuring 6,831 megawatts short of PJM's requirement for the 2026 auction, according to PJM's 2026 auction report.

PJM's independent market monitor tested what would have happened without existing and forecast data center demand. It estimated that including that load increased capacity market revenue for 2026 to 2027 by about $7.27 billion, or 82.1%.

While this is strong attribution data, it doesn't mean that household utility bills increased by $7.27 billion. Utilities use hedging strategies, enter into long-term contracts, and state regulators determine how much of the cost they pass on to each customer. Demand may be the "spark" for these large price increases, but it can only go so far.

This is not only an American argument

The International Energy Agency estimates that data centers consumed 415 terawatt hours worldwide in 2024, about 1.5% of global electricity. It expects consumption to exceed 945 terawatt hours by 2030. That is slightly above Japan's current annual electricity use.

Yet, the global share can be misleading because facilities cluster around available land, fiber connections, favorable regulation, and cheap power. Local effects can be far larger than the worldwide percentage suggests.

Scotland shows how the same facility can be an asset or a liability. The country often pays wind farms to reduce output because the network cannot transport all the electricity being generated. Properly located data centers could consume some of that surplus and reduce the need for expensive transmission construction.

An analysis by the National Energy System Operator presented to Holyrood suggested that locating five gigawatts of data center demand in Scotland could avoid up to £5 billion in network infrastructure over 40 years.

The difficulty arises when the wind is not blowing. Data centers generally require continuous service, so that extra demand may bring gas generation online during periods of low renewable output. Dr. Wei Sun of the University of Edinburgh put the issue plainly during the Scottish Parliament hearing: "Does the data center also pay for that? I think that's really the question we need to answer."

D. Menico Ferraro, an economist at Arizona State University, reached a similar conclusion on the Powering America podcast: "There are signs of strained grids and regional electricity markets that tell us that we are about to enter a new era."

His broader point was that AI may create considerable economic value, but capturing that value requires equipment, skills, and electricity. The argument is therefore about who incurs the cost, not whether a cost exists.

A debate vulnerable to distortion

We need to be careful not to reduce the rising community opposition to data centers to concerns about their electricity bills. Residents have many concerns about water, air pollution, noise, land use, and the limited number of permanent jobs created after construction ends.

Could they be left with nothing but the low humming sound of data centers and low employment? Headlines about a rise in surveillance centers and claims that protesters are engaging in anti-tech extremism suggest there are many reasons to be concerned.

The absence of information is turning an engineering decision into a question of trust. People are being asked to accept the costs of having infrastructure near their homes, even though these costs are very much a local issue. The problem is that many of the promised benefits are being discussed nationally.

The energy debate needs more transparency and less polarization

It's easy to see why residents are angry. Rising prices, combined with constant low-frequency humming and buzzing noise pollution from 24/7 cooling fans and chillers that can reach 50 to 70 decibels, are driving away local wildlife, including birds.

As the AI arms race continues, these real concerns are low-hanging fruit for nation-states that want to manipulate public opinion. In June, OpenAI reported banning a likely China-based cluster that used its tools to generate social media posts criticizing US data centers and AI infrastructure. Sure, the finding deserves attention, but it does not invalidate or distract from genuine complaints.

Grid operators could easily improve transparency by publishing energy use, peak demand, credible connection forecasts, and the infrastructure costs assigned to large energy customers.

AI needs electricity, and countries that cannot provide reliable power will struggle to compete. But this does not give private technology companies an automatic claim on subsidized infrastructure.

The public argument is often framed as a choice between technological progress and protecting households. Maybe we need a less polarizing narrative in which those pursuing the largest rewards step up and carry the financial risk of supplying the power they need.