NextFin News - As politicians in Washington and state capitals move to slow the construction of artificial intelligence data centers, Representative Rich McCormick is warning that the United States cannot afford to hit the brakes. The Republican from Georgia, who chairs the Investigations and Oversight Subcommittee of the House Science, Space, and Technology Committee, argues that the political backlash against the AI infrastructure buildout risks ceding the race for computing capacity to China at the moment when electricity supply, not chip design, has become the binding constraint on American AI leadership.
The warning lands at a hinge point for the trade. A federal moratorium bill would pause construction of new data centers until Congress writes national safeguards. Texas, the largest hub for new projects, has frozen grid approvals pending a statewide audit. Pennsylvania has imposed the strictest state-level standards in the country. And yet power demand from data centers is projected to more than double in two years. The central question is no longer whether AI needs electricity - it is whether American politics will let the grid be built fast enough to supply it.
The Political Backlash Arrives at the Grid
The buildout that defined the AI bull market is running into a wall of local opposition and legislative action. The Artificial Intelligence Data Center Moratorium Act, introduced in the Senate by Bernie Sanders of Vermont and carried in the House by Alexandria Ocasio-Cortez of New York, would impose an immediate federal halt on the construction of new data centers and the expansion of existing ones until "strong national safeguards" are in place. The bill targets facilities with peak power loads in excess of 20 megawatts that use high-performance server racks or liquid cooling, as well as sites used to develop or operate AI models at scale. The House companion drew additional support in late June when Representative Andre Carson of Indiana joined the effort, arguing that families already facing high grocery, gas, and housing costs should not be forced to pay the burden of the technology.
The federal bill has relatively little traction on Capitol Hill, especially among Republicans and Democrats who reject slowing development amid concerns about ceding AI leadership to China, according to policy analysts. But the state-level pressure is real and it is spreading. At least 15 states have weighed pauses on data center development, and at least 100 localities have already approved their own moratoriums. The most consequential action came on August 3, when Texas Governor Greg Abbott ordered the Public Utility Commission of Texas and the Electric Reliability Council of Texas to conduct a "comprehensive verification and audit" of every data center project advancing through ERCOT's interconnection queue before any further approvals can proceed.
"Our top priority is to protect Texans' safety and quality of life," Abbott said in a statement. "Any project that fails to comply with the requirements set forth by the PUCT and ERCOT, and by state law, must be denied connection to the Texas grid. Simply put, Texans must come first."
The scale of what is at stake in Texas is difficult to overstate. ERCOT is tracking more than 1,800 projects in its interconnection queue, representing more than 474 gigawatts of electricity - more than five times the grid's record for peak demand. Roughly 90% of those new power requests are for data centers. Pennsylvania followed on August 18, when Governor Josh Shapiro signed Executive Order 2026-05, which removes all data centers from the state's Fast Track permit program, bars agencies under his jurisdiction from signing nondisclosure agreements with developers, and subjects every proposal to new guardrails on energy affordability, environmental protection, and community engagement.
Against that backdrop, McCormick's argument is that the United States is regulating itself into a position of weakness. In opening remarks at a February hearing on AI data center infrastructure, he laid out the arithmetic plainly.
"Artificial intelligence requires enormous computing power, and that means enormous amounts of electricity. That is not up for debate," McCormick said. "What is important to discuss is whether our approval processes to build infrastructure can meet that demand within a meaningful timeframe. Right now, they are not - we need to understand why and what Congress can do about it."
The Numbers Behind the Constraint
The timing of the political pushback collides with the steepest demand curve the power sector has faced in decades. Goldman Sachs Research projects U.S. data center power demand to climb from 31 gigawatts in 2025 to 41 GW in 2026 and 66 GW in 2027, assuming a 70% capacity utilization rate. The acceleration is the point: year-over-year capacity additions are scheduled to reach 13.6 GW in 2026 and 36.3 GW in 2027, compared with realized additions of just 6.4 GW in 2024 and 8.5 GW in 2025. Data centers' share of total U.S. peak summer power demand is projected to rise from 4.1% in 2025 to 5.3% in 2026 and 8.5% in 2027.
The Energy Information Administration, in its Short-Term Energy Outlook released on August 11, projected that total U.S. power consumption will rise from a record 4,195 billion kilowatt-hours in 2025 to 4,268 billion kWh in 2026 and 4,391 billion kWh in 2027. The agency explicitly tied the surge to data centers dedicated to artificial intelligence and cryptocurrency, alongside the electrification of homes and businesses. In a rare move, the EIA cut its forecast for Texas electricity demand growth in 2027 from 14% to 6% - and named a single cause: the governor's August 3 pause on new data center development. That is politics showing up in the official numbers within days.
On the global stage, the International Energy Agency projects data center electricity consumption worldwide will more than double to roughly 945 terawatt-hours by 2030, with China and the United States accounting for nearly 80% of that growth. The Department of Energy estimates that U.S. data centers consumed about 4.4% of total national electricity in 2023 and could consume between 6.7% and 12% by 2028. These are not marginal loads. They are large enough to move wholesale power prices, reshape regional generation mixes, and determine which technology platforms win the next decade.
McCormick's case rests on the gap between how fast infrastructure can be approved and how fast the technology is moving. Transmission line projects, which are essential to delivering power to new data centers, currently average about four years to permit, and in some cases stretch beyond a decade. Federal environmental reviews under the National Environmental Policy Act averaged 26 months in 2024 - more than two years before construction even begins. Even when agencies ultimately prevail in litigation, which they do roughly three-quarters of the time, the process itself can add years to project timelines.
"We must evaluate whether a permitting framework largely designed in the 1970s is suited to infrastructure that is now central to national security and economic competitiveness in 2026," McCormick said in his February remarks. "That includes reviewing litigation timelines and considering whether AI-related infrastructure warrants differentiated treatment."
Second-Order Effects: The Buildout Does Not Stop, It Migrates
The first-order reading of the moratorium push is simple: fewer data centers get built, and the AI rally loses its footing. The second-order effect is more subtle and more important. The buildout does not stop; it migrates toward the path of least resistance, and the value created by the constraint shifts from developers to whoever controls firm power and permitting advantage.
Consider the market's own positioning. Independent power producers with nuclear fleets and long-term hyperscaler power purchase agreements - Constellation Energy, Vistra, Talen Energy - have become the cleanest expression of the AI trade, because a data center without a guaranteed power supply is just a building. Constellation's chief executive has told investors that projected hyperscaler spending for 2026 is nearly 75% higher than the prior year and continues to be revised upward. Vistra in June formed the Helix Digital Infrastructure joint venture with Nvidia, KKR, and the Kuwait Investment Authority, with $10 billion in capital commitments, designating Vistra as the preferred power provider for new-build and existing projects.
This is the transmission mechanism through which political friction becomes financial value: every additional permitting hurdle raises the option value of capacity that is already permitted, already connected, or politically insulated. A data center in a jurisdiction with a functioning approval process, backed by firm generation that does not depend on new transmission, becomes a scarce asset. The winners are not the companies that promise the most compute; they are the companies that can guarantee electrons on a specific date.
The same logic applies to the chip layer. Nvidia, which supplies the GPUs that make the data centers necessary, traded around $216 in late August ahead of its August 26 earnings report - up only about 2% for the year while the broader market advanced roughly 7%. Wall Street's consensus for the fiscal second quarter calls for revenue near $91.9 billion and earnings per share around $2.08 to $2.09, roughly double the year-ago quarter. The stock's muted performance is itself a signal: investors are beginning to price the possibility that the buildout's rate of change, not its direction, is what matters. A doubling of power demand over two years is bullish for the chipmaker only if the power actually arrives.
The Cyclical Backlash and the Structural Constraint
It is important to separate two forces that are being conflated in the debate. The political backlash is cyclical. It is driven by visible, localized costs - higher electric bills, water use for cooling, noise, land consumption - that land on communities that see little of the revenue. Cyclical pressures reverse: they can be addressed through community benefit agreements, cost-allocation reform, and permitting modernization. McCormick himself has been careful to frame his position this way, saying he is not advocating the elimination of environmental review but arguing that the current process is not meeting its intended goals efficiently.
"I am not advocating that we eliminate all environmental review. We should always be friendly to the environment. We all love to preserve. I'm not advocating that we change the rules in some sort of dramatic capacity. But we do need to change," he said.
The power constraint, by contrast, is structural. Physics and capital cycles do not respond to legislation. A nuclear plant or a large gas facility takes years to permit and build regardless of which party holds the gavels. Transmission lines face the same local opposition dynamics in red states as in blue ones. The grid cannot add capacity at the pace AI demand is arriving, and no moratorium changes that - it only changes where and when the capacity gets built.
This distinction determines the conclusion. If the backlash is cyclical and the constraint is structural, then the buildout continues but slower, more expensive, and more concentrated. The companies that survive the bottleneck are those with firm power, political insulation, and balance sheets long enough to wait out the approval process. The ones that do not are the speculative developers whose projects exist only in interconnection queues - and Texas's 1,800-project queue, 90% of it data centers, is exactly where that culling will happen first.
The Counter-Thesis: Democracy Needs to Catch Up
The strongest case against the buildout is not that AI is a bubble. It is that the decision to reshape the American energy landscape is being made by private companies without democratic input, and that the costs are being socialized while the gains are privatized. Sanders framed it in those terms when he announced the moratorium bill in March.
"We cannot sit back and allow a handful of billionaire Big Tech oligarchs to make decisions that will reshape our economy, our democracy and the future of humanity," Sanders said. "We need serious public debate and democratic oversight over this enormously consequential issue. The time for action is now."
Ocasio-Cortez pointed to concrete harms: surveillance partnerships between immigration authorities and AI companies, deepfakes, and data center construction inflating electric bills in communities across the country. These are not abstract concerns. The EIA's own decision to lower its Texas demand forecast after the governor's pause shows that the political system can, in fact, redirect capital quickly when communities object. A pause gives regulators time to gather the data they currently lack - on water use, on grid reliability, on who ultimately pays for the transmission upgrades.
This counter-thesis has force. But it contains its own falsifying condition. If the moratorium movement gains genuine bipartisan momentum - if the Sanders-Ocasio-Cortez bill attracts meaningful Republican co-sponsorship, or if Texas's audit extends beyond roughly 90 days with zero new approvals granted - then the political risk to the buildout is structural, not cyclical, and the entire premise that permitting reform can unlock capacity needs to be discarded. Until that happens, the evidence points the other way: the federal bill has stalled, Texas's governor remains a supporter of data center investment who is using the audit as a pressure tool rather than a permanent barrier, and the underlying demand curve has not bent.
What to Watch
Three signals will determine whether this is a cyclical detour or a regime change. First, the ERCOT audit: how long it takes, how many projects it clears, and whether the governor follows through with a special legislative session to codify permanent restrictions. Second, the 2027 capacity-additions number: Goldman Sachs Research projects 36.3 GW of new data center capacity coming online next year; if actual additions fall materially below that, the constraint thesis is confirmed and the buildout is slowing for reasons no amount of permitting reform can quickly fix. Third, the equity market's verdict on the power complex - whether nuclear and merchant generators continue to command premium valuations, which would signal that investors see firm capacity as the scarce asset, or whether the theme rolls over, which would signal that the market doubts the demand will materialize at all.
Three scenarios frame the path ahead. The base case is a slower, more expensive buildout: permitting friction persists but does not stop, ERCOT clears most projects after the audit, and Goldman Sachs' 36.3 GW of 2027 capacity additions lands within striking distance of forecast. The upside case is permitting reform: if Congress differentiates AI infrastructure under a modernized framework, the backlog clears, power demand arrives on schedule, and the chip and power complex re-rate higher. The downside case is a political cascade: if the federal moratorium bill gains bipartisan co-sponsorship or Texas's audit hardens into a permanent barrier, projects stall, 2027 additions fall well short of 36.3 GW, and the AI power theme rolls over. Each scenario has a trigger that can be observed, not a forecast that must be believed.
The time horizons point in different directions. In the short term, sentiment around the AI trade is fragile - a disappointing Nvidia print or an extended Texas freeze could trigger a rotation out of the entire complex. Over the medium term, fundamentals dominate: hyperscaler capital expenditure is still rising, and the companies with contracted power will keep building. Over the long term, the structural question is whether the United States can reconcile democratic oversight with the speed that AI infrastructure requires - and whether it can do so without handing the advantage to a competitor that does not face the same constraints.
McCormick's warning is ultimately about that choice. The United States can slow the buildout, but it cannot slow the technology. The question is what happens to a country that decides to regulate its way out of the race while its competitor builds through it.
Explore more exclusive insights at nextfin.ai.
