The data center boom is colliding with the grid’s hardest problems

Electric bills are climbing as utilities confront a widening set of costs and uncertainties.

Across the country, utilities are upgrading aging equipment and hardening systems against wildfires and extreme weather. At the same time, the AI data center buildout is accelerating, with large facilities competing for juice. Serving them can require new generating equipment, transmission expansions, and costly local upgrades – often before anyone knows how much demand will actually show up. Those pressures are compounded by high prices and long lead times for equipment, like gas turbines and transformers.

The scale of the demand growth is still coming into focus, with significant implications not only for climate change but for household budgets. A recent Electric Power Research Institute analysis estimates data centers’ share of U.S. power use could jump from about 4.5 percent today to between 9 and 17 percent by 2030 – a far steeper trajectory than the group projected two years ago. As bills increase, local backlash has intensified, even as the reforms that could ease constraints over time – faster permitting, better transmission planning, smoother interconnection – remain slow and politically fraught.

This was the backdrop for a March 4 meeting President Trump hosted with top tech executives. The White House said the companies signed a “ratepayer protection pledge” meant to keep data center costs from spilling onto household bills. The president also promised to push for faster permitting for related energy projects, though many decisions sit with state and local governments.

To unpack what this arrangement does and doesn’t change, Harvard Climate Brief spoke with Ari Peskoe, director of Harvard Law School’s Electricity Law Initiative, about cost shifting, enforceability, and the surest ways to bring electricity costs down.

What did tech companies commit to in their meeting with President Trump? What is new here, and what would count as proof that it protects household ratepayers?

The White House statement has three pieces: Companies say they’ll cover the cost of new power plants needed for their data centers; they’ll pay for the delivery infrastructure to bring that power to their facilities; and they’ll still pay even if the data centers don’t end up coming online. These terms are, by themselves, not new – companies have been making similar promises for a while. There’s a lot of wiggle room here: Whether the data centers pay for all these costs, or only a portion, can have significant implications for consumers.

Have you seen any binding contracts or enforceable terms? Or is it mostly public commitments that can be reinterpreted later?

This was a press conference and press conferences don’t solve regulatory problems. Nothing becomes real until utilities sign contracts and file them with regulators, or when utilities and market operators file new market designs that actually isolate data center costs. This only becomes real once we have approval from utility regulators.

It’s important to emphasize that utilities hold the pen. Data centers are customers. They don’t set the rates or the terms of service. Utilities and market operators do – and they’re the ones who put the changes on paper and get them approved.

How did we get here? How have data center costs ended up on a regular person’s electric bill?

For the past century, the U.S. model has been that utilities finance, build, and own power plants and power lines, then recover those costs through rates that are spread broadly across customers. This model socializes the costs and made sense when growth was driven by population and the broader economy.

Now a small number of corporations – the largest, wealthiest corporations in the world – are driving billions in new system costs to serve their own facilities. In that context, socialization makes much less sense.

That’s one way households get stuck paying: Utilities build infrastructure and roll those investments into their rates. The second pathway is markets. In regions with competitive wholesale power, prices are driven by supply and demand. Data centers push demand up, supply can’t ramp fast enough, and prices rise – and everyone pays those higher market prices.

The companies agreed to cover generation and transmission costs even if they don’t end up using that power. What does that look like in practice, and how realistic is it?

This may be the most doable part of the pledge. The basic risk for the public is that a utility gears up for a big new data center, spends heavily on new infrastructure, and then the project fizzles – the AI boom cools off, the developer finds a better deal elsewhere – and there’s no data center left to pay those costs. What a number of states are doing, and what utilities often propose, is to require new data centers to sign long-term contracts – often 10 to 15 years – that lock in a payment stream over that period. The big unresolved question is how large that payment should be. States and utilities take different approaches, but we are seeing meaningful steps toward protecting customers from stranded investments.

Describe an example where that’s happening.

AEP Ohio is among the first utilities to move in this direction. It went through a long process that started in May 2024 and produced a new structure for serving large data centers. There are still open questions about whether the companies will pay enough, and whether early financial commitments will lead to better results for consumers. After Ohio regulators approved AEP’s data center tariff, the utility’s projected data center demand dropped from roughly 30 gigawatts to around 13. Presumably, the requirement that data centers sign firm contracts led some speculative ventures to drop out of the queue. That helps protect the public from stranded-asset risk.

But it doesn’t mean households are in the clear. Ohio industrial consumers claim that AEP continues to inflate its data center forecast in order to justify billions in new transmission. Under today’s cost-recovery rules, transmission is paid for by everyone.

Is anyone trying to make data center operators pay for that transmission infrastructure?

Right now, utilities, particularly in the East, are opposed to changing how transmission is paid for. California may be the only state I know that has imposed some requirement on data centers to directly pay for new transmission projects attributable to their demand.

The slow movement from the utility industry on transmission cost allocation comes down to two things. First, once you reopen transmission cost allocation, it’s a big food fight. Everyone wants to protect their own interests. Industrial customers want to pay less, residential consumer advocates want stronger protections, and utilities would rather not relitigate who pays for what.

Second, utilities make money by growing their rate base – the infrastructure they own and earn a regulated return on. If you start directly assigning big transmission costs to a specific customer, that can shift spending into a different bucket where the utility often doesn’t earn the same return. At billion-dollar scale, that’s a significant hit to profit opportunities.

How much of today’s bill increases can credibly be tied to the data center surge versus other major cost drivers? And where are the near-term bottlenecks that could keep prices climbing?

The worry with data centers is that we may be in the early innings of a much bigger buildout – and we pay for new infrastructure for decades. A transmission line is paid for over 20, 30, even 40 years. If we get the rate design wrong now, that mistake compounds for a generation.

That said, there is evidence in some markets that data centers are substantially adding costs for consumers. In PJM [the largest regional transmission organization, serving 13 states and Washington, D.C.], capacity-market prices have surged. The grid operator’s independent market monitor has estimated more than $20 billion in recent costs tied to data center demand. Everybody pays for that. Similarly, that region is also planning massive transmission expansion, largely for data centers: more than $20 billion in the past few years at the regional level. A report from the Union of Concerned Scientists found more than $4 billion in transmission spending across seven states specifically to serve data centers. All these things add up, but they are specific to various states and regions.

What kinds of reforms would lower costs?

Some reforms affect electric bills quickly; others have long-term effects. Take natural gas prices: When they rise, consumer costs move almost immediately, and when they fall, bills can ease.

Improving how we build transmission has a longer-term payoff. If we can cut permitting timelines, that can unlock more development, add capacity to the system, and make it easier for new generation facilities to connect. Those changes can improve outcomes – but over years, not months. The risk is that we lose sight of these reforms and miss a chance to make durable fixes that compound benefits for decades. A lot of the transmission system is ripe for reform, and this is an opportunity to re-evaluate rules that no longer make sense.

– As told to David Trilling