← Front Page
AI Daily
Policy • March 27, 2026

Who Pays for AI's Power? Congress, Ratepayers, and Carbon Credits Enter the Fight

By AI Daily Editorial • March 27, 2026

Three stories converged this week around a single question that the AI boom has so far avoided answering: who is actually paying for all this electricity, and why is it not the companies using it? The answer, increasingly, is that ordinary households and businesses are subsidising the infrastructure build-out for some of the world's most profitable corporations, and the political moment when that becomes untenable may have arrived.

The most direct legislative move came from senators Josh Hawley and Elizabeth Warren, an unusual bipartisan pairing, who urged the Energy Information Administration to establish mandatory annual energy reporting requirements for data centres. The ask is modest on its face: just disclosure, not limits. But disclosure requirements have a history of being the first step toward cost allocation rules, and the context matters. Since 2020, residential electricity prices in the United States have risen by more than 36 percent. In communities hosting large data centres, from rural Virginia to Arizona, local grids are strained and ratepayers are absorbing the infrastructure costs that utilities are not recovering from the tech companies themselves.

The Washington Post reported this week on the growing pushback from communities that hosted data centres with the promise of jobs and tax revenue, and found a different reality. Utilities have in many cases structured their interconnection agreements in ways that spread upgrade costs across all customers rather than charging them to the new large-load customers triggering the upgrades. The political pressure to change this is coming from state utility commissions and consumer advocates rather than Congress, but the Senate reporting push adds federal momentum to what has been a scattered local fight.

Meanwhile, CNBC reported that Big Tech purchases of voluntary carbon credits have exploded, with Microsoft leading by a significant margin. The story frames this as environmental commitment, and the companies involved clearly want it read that way. But the carbon credit market is doing something more interesting than simple offset purchasing: it is providing cover for continued emissions growth while the companies invest in the cleaner infrastructure that may eventually justify their sustainability pledges. Microsoft's carbon credit expenditure is running well ahead of its actual emissions reductions, a gap the company acknowledges is temporary while renewable energy capacity catches up with demand.

What ties these three stories together is a pattern of externalised costs. Data centres are expensive to power, and their power demand is reshaping regional grids, raising prices for households, and generating emissions that are being offset at a cost that does not appear in the price of a ChatGPT query or an Azure inference call. The model has worked for a while because the costs were diffuse and the benefits were visible. The political dynamics are shifting because the costs are becoming less diffuse.

There is a legitimate counter-argument, and it deserves to be taken seriously. Scientific American has argued, with some supporting evidence, that AI applications in climate modelling, grid optimisation, and industrial efficiency could reduce emissions by more than the data centre build-out creates. If that is true, the energy calculus looks different. But that argument depends on the efficiency gains materialising at scale and being captured in emissions reductions rather than rebounded into more compute consumption. Neither of those is guaranteed, and the accounting for it is not yet transparent enough to verify.

The immediate practical story is simpler. Mandatory reporting would force the industry to publish numbers that are currently voluntary or absent. That is not a policy outcome by itself, but it is the precondition for having a real policy debate about cost allocation. The tech industry has been remarkably effective at framing AI infrastructure spending as national economic strategy rather than as private capital investment that relies on public goods. The Senate bill, if it moves, is an attempt to put some numbers on that framing. The carbon credit story is, in its own way, an admission that numbers are going to matter.

Sources