You Are Subsidizing AI's Power Hunger

Your Power Bill Is Not a Coincidence

If your electricity bill has felt punishing lately, there is a concrete reason. US residential electricity prices jumped 17% in 2025 alone, following a 10% cumulative rise between 2022 and 2024. That data comes straight from the US Energy Information Administration. The timing is not random. It maps almost exactly onto the explosion in AI data center construction across the country.

The uncomfortable question this episode of Good Revenue raises is not just why prices are rising. It is who decided that ordinary ratepayers should foot the bill for infrastructure that primarily benefits some of the most valuable companies on earth.

Secret Deals and Invisible Regulators

Utilities are not accustomed to demand shocks. Building new generation capacity typically takes years, sometimes decades, and costs are traditionally allocated to whoever is driving the new demand. That is the logic of the rate base. But AI data centers arrived so fast, and with such enormous appetite, that the system has not caught up.

A 2025 study by a Harvard researcher examined 50 contracts between data centers and utilities and found a troubling pattern. Pricing agreements between tech companies and utilities are largely confidential. In many regulated states, the regulators approving these deals never actually see the price terms. And in states where regulators do have access, there is no consumer advocate at the table to push back on cost allocation. The result is that decisions worth billions of dollars are being made with almost no transparency and no one formally representing your interests.

Neeta Bidwai frames this plainly: we have no reliable way of knowing what any major tech company is actually paying for power in any given market.

How the Tech Giants Are Playing It

The strategies differ significantly by company. Microsoft talks publicly about energy stewardship and has committed to not pushing costs onto local consumers, though without independent verification that pledge is hard to evaluate. OpenAI has been vague in its public statements but explicit about its appetite: it needs 10 gigawatts of new generation capacity by 2029, an enormous ask in a very short window.

Amazon's position is arguably the most revealing. The company has stated publicly that it believes it pays for the electricity it consumes, but that it should not be responsible for grid infrastructure costs like utility poles and transmission equipment. The episode offers a sharp analogy: it is like pumping gas and then telling the station you will only pay for the fuel itself, not refining, distribution, or delivery. That position, however commercially rational it may seem to Amazon, has real consequences for everyone else on the grid.

Google has taken a different path, investing aggressively in vertical integration. Its acquisition of Intersect for just over four billion dollars signals an intent to control more of its own power supply rather than simply negotiate for it.

And here is the twist that complicates the entire picture: utilities are generally prohibited from owning power generation. Tech companies are not. Amazon, Google, Microsoft, and others have been generating electricity and selling it back to utilities at wholesale rates. In recent years, tech companies have sold $2.2 billion worth of electricity back to power companies. That makes them suppliers as well as customers, giving them leverage that ordinary ratepayers simply do not have.

The Grid Is Starting to Crack

The pressure is most visible in the Mid-Atlantic. PJM, the grid operator covering 13 states and the District of Columbia and serving around 67 million people, is home to the highest concentration of data centers in the country. Grid operators there are now warning that AI demand could outpace available generation, raising the real prospect of AI-driven blackouts.

The policy responses are multiplying. The White House wants PJM to change its auction structure. A Maryland senator is pushing legislation to prevent cost increases from being passed to consumers. The Federal Energy Regulatory Commission is being lobbied by tech companies to create a single national standard for adding large power loads, which would streamline the process for them but would, as states have pointed out, risk bypassing the protections built into each state's rate-setting process.

The Bill Is Coming Either Way

The episode closes on a point that deserves to sit with you. Utilities face a two-sided risk. If AI demand materializes as projected and they have not built enough generation, blackouts follow. If they build aggressively and AI demand softens or consolidates, they are left with stranded assets, and ratepayers will almost certainly still absorb those costs.

Either scenario lands on your bill. Without significant Congressional action or meaningful regulatory reform, the current trajectory points one direction. And as this episode makes clear, holding your breath waiting for Washington to act would be ill-advised.