Maine just did what 11 other states tried and failed to do. Its House voted 82-62 to impose a statewide moratorium on new large data centers, halting construction until at least 2027. It is the first state in America to successfully pass such a ban. And the timing could not be worse for the AI industry.

Because Maine is not alone. Tulsa, Oklahoma just unanimously approved its own data center moratorium. Virginia, long the data center capital of the world, is pushing back hard. Analysts now estimate that 30 to 50 percent of the 16 gigawatts of planned US data center capacity scheduled for 2026 will be cancelled or delayed. Only 5 gigawatts is actually under construction right now.

Let those numbers sink in. The AI industry has promised to spend hundreds of billions on infrastructure this year. OpenAI, Microsoft, Google, Meta, and Amazon all have massive data center buildouts planned. And the states they need to build in are telling them to stop.

The reasons are piling up fast. Communities are furious about electricity consumption. Maine legislators cited concerns about water usage, power grid strain, and the fact that data centers create remarkably few jobs for their size. One facility can consume as much electricity as a small city while employing fewer than 50 people. The backlash has turned violent in some areas. A councilman who supported a data center project recently had his house shot up with his 8-year-old inside.

The math problem is straightforward. AI companies need power. They need it fast. And they need a lot of it. But 25% of planned data center projects have not even disclosed a powering strategy. They simply assumed they would figure it out. Now the communities they planned to build in are saying: figure it out somewhere else.

Meanwhile, Congress remains stuck. Bernie Sanders and Alexandria Ocasio-Cortez introduced a federal data center moratorium bill last month. It has gone nowhere. So states are acting on their own, creating a patchwork of local bans that makes long-term infrastructure planning nearly impossible for Big Tech.

Here is the second-order effect nobody is talking about. If half of planned data center capacity gets cancelled or delayed, the compute bottleneck that is already strangling AI development gets dramatically worse. Model training runs that need massive clusters will face longer queues. Cloud pricing goes up. Startups that cannot afford their own infrastructure get squeezed out. The AI boom does not stop, but it slows down in the one country that is supposed to be leading it.

Maine voted to pause and study the problem before letting more data centers in. That sounds reasonable. The AI industry promised jobs and tax revenue that never materialized while draining local power grids. What is unreasonable is expecting communities to absorb those costs indefinitely. The states are not killing AI infrastructure. They are asking a question Big Tech cannot answer: what is in it for us?