Trump seeks to shift AI power costs

President Donald Trump has moved to reshape how the United States pays for the explosive growth in electricity demand driven by artificial intelligence, backing a plan to use 2026 capacity auctions in the PJM Interconnection to force large technology companies to shoulder a greater share of grid expansion costs. The proposal, framed as a way to protect households and small businesses from higher power bills, places AI data centres at the centre of an intensifying debate over energy pricing, infrastructure investment and economic fairness.

The PJM Interconnection, which manages electricity transmission across 13 states and the District of Columbia, is the largest power grid in the country and the primary arena for the proposed change. Under the plan endorsed by the administration, capacity auctions scheduled for 2026 would be structured to reflect the outsized and continuous electricity needs of hyperscale data centres that support AI training and cloud computing. Operators seeking guaranteed access to large volumes of power would face higher, more targeted charges, effectively requiring them to contribute directly to the cost of new generation and transmission infrastructure.

Trump has argued that the rapid expansion of AI has created a structural imbalance in the power market. Data centres, often clustered in regions with relatively cheap electricity, draw enormous and steady loads that require utilities to invest billions of dollars in new power plants, grid upgrades and backup capacity. Those investments, he says, have too often been socialised through higher rates for ordinary consumers rather than borne by the companies driving demand. “If you’re building massive data centres and using unprecedented amounts of electricity, you should be paying for the system that supports you,” he told advisers in recent policy discussions.

Industry analysts estimate that electricity demand from data centres in the United States could more than double by the end of the decade, with AI workloads accounting for a growing share of that increase. Some projections suggest AI-related computing could consume close to a tenth of national electricity output by 2030, a sharp rise from today’s levels. PJM has already warned that it faces a tightening supply-demand balance later this decade as older fossil fuel plants retire and new generation struggles to come online fast enough.

The administration’s approach relies on the mechanics of PJM’s capacity market, where power generators are paid to commit to being available during periods of peak demand. By adjusting auction rules to better reflect the scale and reliability requirements of large data centre loads, regulators could ensure that those entities pay higher clearing prices or additional fees tied to infrastructure build-out. Supporters say this would send clearer price signals and reduce the risk that residential and small commercial customers subsidise the growth of some of the world’s most profitable technology firms.

Utilities have broadly welcomed the principle of aligning costs with demand, though they caution that market design changes must be carefully calibrated. Grid operators note that data centres can provide long-term, predictable demand that supports investment, but only if pricing frameworks are stable. Sudden or poorly designed shifts, they argue, could deter capital or push projects to other regions, complicating long-term planning.

Technology companies, meanwhile, have expressed concern that the proposal could significantly raise operating costs and slow the deployment of AI infrastructure in the United States. Several firms have invested heavily in renewable energy contracts and efficiency measures, arguing that they are already contributing to grid resilience and decarbonisation. Executives warn that higher capacity charges could ultimately be passed on to customers or reduce funds available for innovation.

Environmental groups are divided. Some see the plan as a necessary check on unchecked energy consumption, encouraging more efficient data centre design and greater investment in clean generation. Others fear that making it easier to build new power plants quickly, even if funded by technology firms, could prolong reliance on fossil fuels if renewable and storage projects fail to keep pace.

Within policy circles, the proposal is also viewed through a political lens. By targeting large technology companies, Trump positions the plan as a populist measure aimed at shielding voters from rising living costs while asserting greater control over an industry often criticised for its influence and scale. At the same time, the administration must navigate the complex regulatory landscape governing wholesale power markets, where changes require approval from federal energy regulators and coordination with state authorities.