President Trump is moving to make the companies driving the artificial intelligence boom shoulder the cost of the power plants that keep their data centers running. Instead of letting utilities and households absorb the bill for AI’s soaring electricity demand, the administration wants Big Tech to underwrite a new wave of generation capacity across the Mid-Atlantic grid.
The proposal would tie long term power contracts to new natural gas and potentially nuclear plants, effectively turning AI’s energy hunger into a financing engine for large scale infrastructure. I see it as an attempt to rewrite who pays for the next chapter of the grid, and to decide whether AI becomes a public burden or a private line item for the firms that profit most from it.
The emergency auction that puts AI on the hook
The centerpiece of the plan is an emergency wholesale electricity auction that would force technology companies to lock in years of payments for new power plants, even if they do not end up using every megawatt. President Trump and the governors of several U.S. Northeastern states have agreed to push the region’s grid operator to run this special auction, targeting the same Mid-Atlantic system that already hosts some of the country’s largest data centers. The idea is to turn AI’s projected load growth into guaranteed revenue streams that can justify building large plants quickly.
According to the administration’s own description, the auction would be structured so that new natural gas generation and potentially nuclear facilities can secure long term contracts backed by tech companies’ balance sheets, rather than by ordinary ratepayers. Officials have framed the initiative as a way to “usher in the age of artificial intelligence” with new power plants funded by the technology companies themselves, a goal that is explicitly embedded in the call for an emergency auction from the Trump administration. By tying AI’s growth to concrete infrastructure commitments, the White House is signaling that the era of cheap, consequence free power for data centers is over.
Why the PJM grid is ground zero
The focus on the Mid-Atlantic is not accidental. The White House and governors want the grid operator that manages electricity for a vast swath of the Atlantic seaboard to hold the auction, targeting the same network that already faces price spikes and reliability concerns as AI data centers proliferate. That operator, known as PJM, runs the largest electric grid in the country, stretching from Illinois to New Jersey and serving more than 65 million people, which makes it the logical test bed for a policy that could reshape how digital infrastructure is powered.
In practical terms, the move would effectively force companies to help fund new power plants in the PJM region as soaring demand from AI data centers pushes up electricity use and wholesale prices. Reporting on the proposal makes clear that the administration expects Big Tech to sign contracts worth billions of dollars to support new capacity in PJM, rather than relying on traditional rate based utility investments. When I look at the geography and the stakes, it is obvious why this grid is ground zero: it combines dense clusters of cloud campuses with aging plants and a market design that already leans heavily on forward capacity auctions.
Big Tech’s $15 billion headache
For the companies that have spent years touting carbon free energy pledges, the numbers attached to this plan are jarring. The Trump administration wants tech companies to buy $15 billion of power plants they may not use, a figure that underscores how much new steel in the ground officials believe is needed to keep AI servers humming. In effect, the government is telling hyperscalers that if they want to keep training ever larger models, they must also become anchor customers for a fleet of new gas and nuclear units.
One analysis of the proposal notes that these obligations could be locked in through long term contracts that resemble insurance policies for the grid, with tech firms paying for capacity that might sit idle during normal conditions but is available during peaks or emergencies. The description of $15 billion in potential commitments, reported by Tim De Chant at 11:38 AM PST, captures the scale of what is being asked. From my vantage point, this is less a nudge than a hard shove, designed to convert AI’s abstract “energy hunger” into a line item that chief financial officers can no longer ignore.
Shifting costs off households and onto data centers
Behind the technical language of auctions and capacity markets sits a blunt political calculation: voters are already feeling the impact of higher power prices, and the administration wants to show it is protecting households from AI driven spikes. President Trump and the Northeastern governors have framed the initiative as a way to stop ordinary customers from subsidizing the electricity appetite of cloud giants, especially in states where data center clusters have grown faster than local generation. By insisting that the companies driving demand sign up for long term contracts, they are trying to shift the financial risk of new plants away from ratepayers.
Several reports describe how the White House has been alarmed by AI related power shortages and price surges in the Mid-Atlantic, with officials warning that without new capacity, reliability could deteriorate as more data centers connect to the grid. The White House and Atlantic state leaders have jointly pressed the grid operator to design an auction that directly targets the tech sector, a push that has been detailed in coverage of power shortages and AI driven demand. When I connect those dots, the strategy looks like a classic attempt to align political incentives with market design: make the most profitable players pay, and hope that voters see lower bills and fewer blackouts as a result.
The broader fight over AI, climate and control
There is also a deeper policy clash embedded in this plan, one that pits the administration’s preference for large gas and nuclear plants against tech companies’ public commitments to renewable energy. The emergency auction is explicitly designed to fast track the development of natural gas generation and potentially nuclear plants by guaranteeing revenues, according to officials who have briefed on the proposal. That approach is consistent with earlier moves by President Trump to favor firm, dispatchable power over intermittent wind and solar, and it now collides directly with AI companies’ efforts to brand their data centers as green.
At the same time, the Trump administration and a bipartisan group of governors are using this moment to push for broader reforms in how the largest electric grid in the country plans for and pays for capacity. Their call for changes to PJM’s market rules, described in detail in coverage from Washington, suggests that AI is becoming a lever to revisit long standing debates over who controls the grid’s future. When I step back, I see more than a fight over one auction: it is a struggle over whether the next wave of AI infrastructure will be built on the back of public utilities or on the balance sheets of the very companies that made artificial intelligence the hottest technology of the decade.
More from Morning Overview