Trump’s “Data Center Power Tax” Is Coming — And It Could Blow Up Big Tech’s AI Cost Model

Trump is gearing up to take a very direct swing at the political problem that keeps showing up on utility bills: data centers. The Wall Street Journal that the administration is expected to unveil an unprecedented plan aimed at PJM Interconnection (the largest U.S. grid operator, spanning 13 states and serving roughly 67 million people) that would shift more of the cost of new power generation onto the tech companies building electricity-hungry AI infrastructure. The core idea: an “emergency” auction where tech companies effectively fund new power plants

According to reporting on the proposal, the Trump administration and a bipartisan group of governors plan to push PJM to run a one-time “reliability backstop” style auction in which tech companies (i.e., data center owners/operators) would bid for 15-year contracts tied to new generation capacity. Those contracts could be worth billions and are being framed as a way to get steel in the ground faster by giving developers long-duration revenue certainty—something power markets don’t reliably provide. The White House has characterized the approach as a way to fund roughly $15 billion of new generation in PJM territory, with an emphasis on preventing those costs from landing on households and small businesses.

This initiative is also closely tied to PJM’s recent capacity-market dysfunction: prices in recent auctions have been extremely high (hitting caps) and, in at least one case, the market didn’t procure the targeted surplus capacity—exactly the kind of headline that turns “grid reliability” into “why is my bill up again?”

The administration’s messaging is blunt: it does not want voters paying higher electricity bills “because of data centers,” and it wants Big Tech to “pay its own way.” That framing is tailor-made for a cost-of-living cycle, especially in PJM states where capacity costs have become a flashpoint for governors and regulators.

There’s also a practical policy logic underneath the politics. PJM is the epicenter of U.S. data center density (especially Northern Virginia), and it’s dealing with a collision of (1) accelerating load growth from AI/data centers, (2) power plant retirements, and (3) slow interconnection/permitting timelines. Even when auction prices scream “build,” you can’t conjure a gas turbine (or a nuclear plant) out of thin air—lead times and supply chains don’t care about your election calendar.

How this hHts Hyperscalers: Sosts shift from “Rate Base” to “Checkbook”

For hyperscalers (Amazon, Microsoft, Google, Meta, etc.), the big change is philosophical and financial: electricity stops being a pass-through line item and starts looking like a semi-fixed capacity obligation. A 15-year contract where you pay whether you consume the electrons or not is basically “take-or-pay” for power—great for generators, less cute for CFOs. The largest platforms are more capable of absorbing that kind of commitment because they have scale, financing flexibility, and (often) pricing power to pass costs through to cloud customers. The more awkward squeeze is likely for smaller AI infrastructure players or colocation providers that don’t have the same margin cushion or contract structure—higher power costs can hit them like a tax on growth.

There’s a second-order benefit for the hyperscalers, though: reliability. If PJM is genuinely short on capacity and interconnection queues are jammed, a system that forces “serious bidders only” could reduce speculative data center requests and improve demand forecasting. In other words, the plan is partly an affordability move, and partly a “stop clogging the pipe with maybes” move.

Which Energy Companies Could Be Impacted: Follow the Money to Whoever Builds

If the auction structure does what it’s designed to do—pull forward construction—then the main beneficiaries are the developers and suppliers of firm generation and the infrastructure around it.

Likely winners by category (not a promise, just the direction of travel):

• Independent power producers and developers that can build in PJM and monetize 15-year contracted cash flows (a big upgrade versus volatile merchant economics)

• Natural gas ecosystem: turbine makers, engineering/construction firms, and gas pipeline/fuel suppliers, because gas remains the fastest path to scalable, dispatchable capacity in the near term

• Nuclear (longer-dated): the plan explicitly raises the possibility of “guaranteed revenues,” which is exactly the kind of structure nuclear developers want—but timelines are longer and permitting is harder, so near-term impact is more narrative than megawatts

Meanwhile, regulated utilities could see a different kind of impact: if more of the incremental build is “data center-funded,” it reduces political pressure to push massive costs into general ratepayer bills. That’s good for consumer optics, but it can complicate the traditional utility playbook of building assets into rate base.

The Consumer and Grid Impact: Relief… with an Asterisk the Size of a Substation

If executed cleanly, the consumer-facing pitch is straightforward: households and small businesses are less exposed to the AI buildout driving capacity costs. That’s especially relevant because PJM capacity charges are a real bill item, and they’ve been hitting record levels.

But there are trade-offs:

• Somebody still pays. Shifting costs to data centers could protect residential bills at the margin, but it raises the cost base of AI infrastructure—which eventually flows into cloud pricing, enterprise software costs, and “why is everything a subscription now?” economics.

• It may advantage the biggest players. Long-term obligations and big bids tend to favor hyperscalers over smaller entrants, potentially reinforcing concentration in AI compute.

• Execution risk is non-trivial. PJM isn’t even publicly celebrating this plan yet (reporting suggests the grid operator wasn’t invited to the announcement), and anything that touches wholesale market design tends to end up in stakeholder fights and, eventually, FERC proceedings. (ZeroHedge)

The Permitting Angle: EPA and Grid Hookups are Part of the same Chessboard

Finally, this isn’t just about auctions—it’s also about building faster. Alongside the PJM concept, the administration is also moving on permitting constraints. The EPA has proposed changes around Clean Water Act Section 401 that would limit states’ and tribes’ ability to block major energy projects, arguing the process has been used to stop projects for political reasons rather than water quality. If that effort survives legal and regulatory scrutiny, it could reduce friction for pipelines, transmission, and other infrastructure needed to support data center load growth.

Bottom Line

Trump’s emerging “data centers pay their fair share” push looks like a targeted attempt to cap the political fallout of AI-driven power demand by forcing a new funding model for generation in PJM—one that prioritizes long-term contracts, faster build incentives, and ratepayer protection. For hyperscalers, it’s a shift from variable power pricing to capacity-backed commitments (more certainty, more obligation). For the grid, it’s potentially constructive—if it accelerates firm supply and unclogs speculative demand. For consumers, it’s marketed as bill protection. For everyone else, it’s a reminder that the AI boom runs on chips, yes—but also on permitting, turbines, and the unglamorous art of keeping the lights on.