Local opposition blocked $68B of US data centers in one quarter
Two hard numbers landed today on the two most-hyped physical bets in AI — the buildings that train the models and the robots meant to work in them. Both came in far smaller than the money already committed to them.
Local opposition blocked or delayed 45 US data center projects worth $68 billion between April and June — more than half of every large-scale development the research group Data Center Watch tracked in the quarter. The figures, provided to Bloomberg, show a backlash that has hardened into a national political force: 843 opposition groups now operate across 49 states, with Hawaii the only exception, and roughly 30 statehouses introduced or adopted rules covering data center siting, electricity and water. Communities are increasingly passing moratoriums on new construction before developers can even file for permits.
That mechanism is what makes the money number read wrong. A blocked project is not a delayed one, and the quarterly totals swing hard — the first three months of the year saw 75 cases worth about $130 billion, against 45 worth $68 billion in the second. One petition against a Tennessee project collected more than 500,000 signatures, over a third of everything signed on Change.org that quarter. Data Center Watch also logged opposition outside the US for the first time, citing campaigns in Europe, Australia and South Africa; lead analyst Miquel Vila said the phenomenon "is increasingly becoming global." We have tracked this as a tax on the buildout in The data center backlash is the new NIMBY tax on AI, and the national mood we reported in 61% of Americans now oppose local data centers is the same signal at a bigger sample. What changed is the instrument: opposition is no longer a public-comment problem that a better community-benefits package can absorb, it is a queue problem — and queues formed before an application exists cannot be negotiated with.
The International Federation of Robotics counted roughly 7,000 humanoid robots sold worldwide in 2025 for industrial and professional use, and says many of them never did productive work. IFR Secretary General Susanne Bieller told Reuters the units account for "only a small fraction" of global robot sales, and that many were bought by research institutes and companies collecting data to improve AI models rather than to fill a job. The tally excludes household and military machines, and counts anything that resembles a human and operates autonomously in environments designed for people — legs are optional. Seven thousand units is the annual volume of a niche industrial product line, not a platform shift, and it arrives after a year of capital that priced humanoids as the next general-purpose computer.
What to watch: whether Q3's opposition number bounces back as the moratorium wave spreads, and whether IFR's next tally splits research purchases from deployed work — that split, not the unit count, is the honest measure of the humanoid market.
Which of these two bets would you underwrite today — the buildings or the robots? Tell us in the comments.
