NextFin News - President Donald Trump has issued a blunt warning to the technology industry: America's artificial intelligence data center boom will not be paid for by American households. As electricity demand from data centers races to more than double by 2030, the president is telling hyperscalers to build their own power plants, fund their own grid upgrades, and pay for the electricity whether they use it or not — or face a political backlash that is already spreading through state capitols from New York to Texas.
The tension at the heart of the moment is stark. The White House is simultaneously the AI industry's biggest cheerleader and its sternest bill collector. Trump has called data centers "bigger than oil" and "Money Machines," while also warning that an "old grid" that "could never handle the kind of numbers" of electricity the AI buildout requires means tech companies must "produce their own electricity." The question investors should be asking is not whether the warning is real — it is — but whether a voluntary pledge can hold when the economics of the grid say someone has to pay for the wires, the plants, and the risk either way.
The Warning and the Pledge
The administration's answer is the Ratepayer Protection Pledge, first unveiled in March 2026 and expanded in July, under which leading US hyperscalers and AI companies agree to "build, bring, or buy all of the energy needed for building and operating data centers, paying the full cost of their energy and infrastructure, no matter what." The pledge carries five commitments: build or secure new power supply; pay for new power delivery infrastructure upgrades; pay whether they use the power or not; invest in local jobs and workforce development; and contribute to electric and community resilience.
"We have an old grid, it could never handle the kind of numbers — the amount of electricity — that's needed. So I'm telling [companies] they can build their own plant; they're going to produce their own electricity," Trump said as he laid out the new ground rule for Big Tech's AI buildout.
Major tech companies have signed on. Microsoft President Brad Smith wrote that "it's both unfair and politically unrealistic for our industry to ask the public to shoulder added electricity costs for AI," committing the company to "pay their own way for the electricity costs they create." OpenAI's chief operating officer pledged "paying our own way on energy so our operations don't raise electricity bills for local residents," and Anthropic said it would "pay for 100% of the grid upgrades needed to interconnect our data centers, paid through increases to our monthly electricity charges."
But the pledge is voluntary, and the politics are moving faster than the signatures. New York Governor Kathy Hochul signed an executive order imposing up to a one-year moratorium on large data centers using 50 megawatts or more. Texas, a Republican state and the epicenter of the gas-fired buildout, moved to tighten regulation of data center construction. Bills to ban data center construction have been introduced in 15 states, and Data Center Watch estimates that roughly $98 billion of projects were blocked or delayed in the second quarter of 2025 alone.
Trump has dismissed the opposition as a public relations problem. "They need some PR help, because people think that if a data center goes in there, electricity prices are going to go up," he said. Polling suggests the opposition runs deeper than PR: an Ipsos poll in June found Americans opposed new data centers 45% to 16%, with opposition rising to 59% when the facility would be within 10 miles of home; a Gallup poll in March showed seven in 10 Americans opposed building data centers in their area; and a May Marquette Law School poll found a 71% to 29% margin that the costs outweigh the benefits.
Why the Grid Math Does Not Add Up for Ratepayers
The scale of the demand shock is what makes the payment question unavoidable. The International Energy Agency projects global data center electricity consumption will more than double to about 945 terawatt-hours by 2030 — slightly more than Japan's entire electricity consumption today — and climb to around 1,200 TWh by 2035. From 2024 to 2030, data center demand is growing at roughly 15% a year, more than four times faster than electricity consumption across all other sectors. The United States accounts for the largest share of that increase: US data center consumption is set to rise by about 240 TWh, up 130% from the 2024 level, with American data centers responsible for nearly half of US electricity demand growth over the decade. China and the United States together account for nearly 80% of global growth through 2030.
That demand is arriving on a grid that was not built for it, and the bill for rebuilding it is large. The Edison Electric Institute estimates utilities will invest $1.4 trillion in 2026-30, up $300 billion from the prior decade's plan. Morningstar expects utility capital investment to climb 17% in 2026 and 10% in 2027. Goldman Sachs Research forecast in February 2026 that electricity prices would rise 6% through 2026 and another 3% in 2028 as data center demand grows more rapidly than power supply.
Households are already feeling it. Bureau of Labor Statistics data showed electricity prices up 6.7% over the 12 months to January 2026, and Morningstar reported average US residential electricity prices rose nearly 8% in the first quarter of 2026 from a year earlier. Trump's campaign promise to cut electricity bills in half has run directly into the AI buildout — and 77% of respondents in the June Ipsos poll said data centers were making electricity more expensive.
Yet the causal chain is not as simple as "data center arrives, bills rise." A Lawrence Berkeley National Laboratory study found that serving data centers is not the main driver of recent cost increases; states where load has risen have generally seen price reductions, though households did not necessarily see lower costs. The bigger cost drivers are spending on grid infrastructure, transmission and distribution upgrades, disaster recovery, and natural gas price variability. Stanford's Bits & Watts Initiative director Liang Min put it plainly: "With or without data centers, we need to harden our infrastructure. We have to build new transmission and distribution equipment. Bills would go up even without data centers."
This is the crux of the matter. The grid needs rebuilding regardless of AI. The data center boom simply accelerates the timetable and concentrates the cost. The pledge's most important clause may be the third one: pay whether they use the power or not. That is an attempt to make data centers behave like firm, contracted load — to ensure the plants and wires get built even when the servers sit idle — and to stop utilities from socializing the fixed cost of capacity across households that never send a single query to a chatbot.
The Second-Order Trade: Who Wins and Who Pays
The first-order effect of the pledge is obvious: if tech companies truly fund their own generation and interconnection, ratepayers are insulated from the marginal cost of AI demand. The second-order effect is more consequential for investors. Capital will rotate toward whoever owns the electrons and the wires, and away from whoever is exposed to political risk.
Utilities with constructive regulatory environments stand to benefit. Morningstar analysts identified Entergy, Pinnacle West, WEC Energy Group, and Southern as among the best-positioned to profit from data center growth, and viewed the sector as roughly 5% undervalued as of May 1, 2026. The logic is that data center expansion adds a new, large, creditworthy source of demand growth that has not yet been fully priced into utility stocks. Independent power producers and gas generators in Texas — where 39 proposed gas-fired plants, nearly half the national total, are being developed to serve data centers — sit on the other side of the same trade.
The market, however, is not yet convinced. The utilities sector ETF, XLU, was up just 1.43% year-to-date as of late August 2026, compared with 12.44% for its category — a sign that investors are treating the data center demand story as something utilities must execute through, not a free option. The sector's lagging performance captures the central uncertainty: more demand is good for volume, but only if regulators allow the cost to be recovered and if communities allow the infrastructure to be built.
But there is a catch that the pledge does not resolve. Data centers still need the grid. Even if a company builds its own power plant, it needs transmission to reach the site, distribution to move power inside the campus, and backup capacity for when its plant trips. Those costs flow through utility rate bases unless regulators carve out data centers entirely — and every carve-out shifts cost onto the remaining customers, which is exactly the outcome the pledge is meant to prevent. The mechanism is circular: the more successfully data centers are walled off from the grid's fixed costs, the more those fixed costs concentrate on households and smaller businesses, and the hotter the political backlash becomes.
That political risk is now a priced factor. On August 26, 2026, the president declared a national emergency to secure the United States bulk-power system against foreign equipment, underscoring how national security, grid reliability, and AI infrastructure have fused into a single policy bundle. The administration's own energy secretary warned in March that if tech companies "are perceived to drive up electricity prices," they will reap the backlash. In other words, the pledge is not just an economic arrangement; it is political cover.
The Counter-Thesis: This Is a Cyclical Backlash, Not a Structural Brake
The strongest argument against reading Trump's warning as a durable constraint on the AI buildout is that the opposition is cyclical, not structural. Public anger over electricity bills is real, but data centers deliver jobs, tax revenue, and capital investment to communities that often need them. Trump's own framing — "Cash Cows, with Lower Taxes and Record Setting Jobs," and "any community that turns down AI is making a big mistake" — is a bet that the economic benefits will eventually outweigh the sticker shock. History supports the cyclical view: previous infrastructure buildouts, from interstate highways to fiber optics, faced fierce local resistance before becoming politically untouchable. If the grid additions get built and bills stabilize, the backlash fades.
There is also a technological counter-argument. Chip efficiency keeps improving, and AI inference costs keep falling. If the energy intensity per query drops fast enough, the demand curve could bend before the political curve breaks. The IEA's projections already assume efficiency gains and still forecast a doubling of demand — but the uncertainty band around 2030 is wide, and a faster-than-expected efficiency breakthrough would be the single biggest bear case for the data center power thesis.
The counter-thesis fails on one point, however: timing. Even if the backlash is cyclical, the grid buildout is not. Transmission lines take a decade or more to permit and construct; gas plants and nuclear projects take years. The demand is arriving now, and the supply response is structurally slow. That mismatch — not the politics — is what will determine prices and profits over the next five years. The politics merely decide who writes the check in the interim.
What to Watch
Three signals will tell investors whether the pledge is holding or fraying. First, electricity price inflation: if the BLS measure of electricity prices prints above 6% year-over-year for two consecutive quarters through 2026-27, the political pressure will intensify regardless of how many companies have signed the pledge. Second, state-level action: more than a dozen moratorium and siting bills are pending, and a wave of approvals in red states like Texas would signal that the backlash is contained regionally even as it grows nationally. Third, the capex numbers: if utility capital spending for 2026 comes in materially below the projected 17% increase, it would suggest developers are struggling to finance or permit the buildout fast enough to meet demand.
The base case is that the pledge holds for the largest hyperscalers — Microsoft, OpenAI, Oracle, and their peers have both the balance sheets and the political incentive to comply — while smaller developers and colocation providers get squeezed by higher interconnection costs and local opposition. The upside case for the AI buildout is that efficiency gains and new nuclear and gas capacity arrive faster than expected, easing the price pressure and defusing the politics. The downside case is a feedback loop: bills keep rising, states impose moratoriums, projects are delayed, and the US cedes data center leadership to regions with more permissive siting regimes.
Split by horizon, the picture is mixed. In the short term, sentiment and polling will drive volatility in utility and data center stocks. Over the medium term, fundamentals will favor whoever owns contracted generation and transmission in data center hubs. Over the long term, the structural question is whether AI's electricity demand proves to be a once-in-a-generation grid buildout — the kind that creates durable utility franchises — or a cyclical capex spike that leaves stranded assets when efficiency catches up.
The bottom line: Trump's warning is less a threat to the AI boom than a recognition that the boom's bill has come due. Data centers may well be bigger than oil in their economic promise, but unlike oil, they cannot be drilled anywhere — they need a grid, a permit, and a community that will tolerate them. The companies that bring their own power and their own political capital will build the future; the ones that assumed the grid would absorb the cost are about to learn that the ratepayer protection era has begun.
Explore more exclusive insights at nextfin.ai.
