Generated by Codex with GPT 5.6 Sol XHigh

This summary covers The Economist’s August 22nd, 2026 Business/Schumpeter article listed in the contents as Data-centre politics and published under the headline Shadow boxing.

America’s backlash against data centres looks fiercer than it is. Politicians are denouncing speculative projects, imposing reviews and announcing restrictions as voters grow anxious about noise, pollution and electricity use. Yet many of the projects being attacked were never likely to be built, while rules aimed at the industry often spare developments already under way. The article’s central argument is that developers and politicians are both exaggerating: one side inflates its construction plans, and the other inflates its power to stop them.

Vapourware meets electoral politics

Kevin O’Leary’s proposed data-centre complex in Utah illustrates the pattern. The television personality announced what he called one of the world’s largest such projects without explaining how it would be financed or powered. An analyst dismissed it as “vapourware gigawatts”. Even so, the proposal became a potent local issue. Republican politicians who had supported it demanded that its size be cut, and several lost primary elections amid the controversy.

Data centres create an awkward political coalition. Many voters see them as ugly, noisy and power-hungry monuments to Silicon Valley, but markets are counting on enormous artificial-intelligence infrastructure investment. Organised labour often supports the construction work, leaving Democrats caught between local opposition and unions. Politicians therefore have an incentive to appear tough without stopping the projects most likely to create jobs and attract capital.

That helps explain Pennsylvania governor Josh Shapiro’s campaign against “speculative proposals”. Of more than 100 announced projects in the state, fewer than one-fifth had sought the permits needed to proceed. The distinction matters because the announced pipeline mixes credible plans from Google and Amazon with far less certain proposals from property developers, private-capital firms, crypto-miners and new cloud operators. Bernstein analysts count the technology giants’ promised capacity in full, discount plans from CoreWeave and Nebius by about half, and assign no weight to O’Leary’s project.

Other supposed cancellations are equally misleading. Vermaland announced a 3-gigawatt Arizona facility, then reportedly reduced it by four-fifths after local opposition. A separate 1-gigawatt proposal in Florida was shelved after an obscure investment firm had bought the land for just \$15m. When a project that was unlikely to materialise is cancelled, the political noise can be much larger than the economic effect.

Restrictions with escape hatches

Several high-profile policies are softer than their headlines. New York announced a one-year moratorium on large projects, but exempted data centres already approved and is not central to the national construction boom. Virginia imposed a \$600m electricity tax while retaining a sales-tax exemption worth roughly three times as much. Pennsylvania’s new process requires local approval before state review, allowing the governor to delay politically awkward decisions until after the election.

Texas offers the clearest test. Its grid operator has received requests for 474 gigawatts of new connections, more than five times the system’s peak demand of roughly 90 gigawatts. Governor Greg Abbott announced an audit of data-centre plans that is expected to finish just after the midterm elections. Some of the industry’s largest operators support the review, suggesting that it may clear away less credible competitors rather than halt serious developments.

Political conflict will delay some projects, especially those that have not broken ground. But the share of the American data-centre pipeline delayed or cancelled has held at about 10% over the past year, even as announced capacity and public opposition have surged. Developers remain more constrained by chips, skilled workers and other market bottlenecks than by government. If one location becomes hostile, they can move to friendlier western states, federal land or eventually other countries.

The article contrasts China’s ability to build by command with America’s habit of willing projects into existence through exaggeration. Developers announce more than they can deliver; politicians promise to stop more than they are willing or able to block. Those claims often cancel each other out. The real data-centre boom may be smaller than the industry’s press releases suggest, but it is also more durable than the political backlash implies.