AI's $10.3T Buildout Faces Massive Local Pushback
The biggest infrastructure project in US history is being stalled by angry neighbors and local zoning boards.
Business · Source: Traders Union (FT syndication)
What happened
The AI industry plans to spend $10.3 trillion on data centers by 2032. This dwarfs the construction of the interstate highway system and the railroads. But the money is hitting a brick wall. In the first six months of 2026 alone, local opposition blocked or delayed 120 data center projects worth $198 billion.
Residents are fighting back over power bills, water usage, and environmental impact. A recent Pew Research survey shows 54 percent of Americans now view data centers as bad for the environment. Half worry these facilities will spike their home energy costs. In Windsor, Colorado, a proposed data center drew 300 angry residents to a town meeting. A typical commercial project there usually draws about thirty.
Politicians are taking notice of the backlash. New York just slapped a one-year ban on new data centers requiring 50 megawatts or more. California passed new AI reform laws. This November, voters will decide on 35 different ballot measures targeting data centers across states like Florida, Kansas, and Ohio. The industry claims they provide a million jobs and billions in taxes, but academic studies show rural towns often lose out on net benefits.
Key facts
- $10.3 trillion — Estimated value of the data center buildout through 2032
- 120 — Number of data center projects blocked or delayed in the first half of 2026
- $198 billion — Value of the stalled data center projects
- 54 percent — Share of Americans who consider data centers bad for the environment
- 50 megawatts — Power threshold triggering New York's one-year ban on new data centers
Why it matters
If you build AI products, your compute costs are directly tied to this physical infrastructure. The cloud feels infinite, but it runs on concrete, steel, and local power grids. When 120 mega-projects get delayed, the supply of compute tightens. That means higher API costs and fewer available GPUs for startups. You cannot scale software if the hardware has nowhere to plug in.
The narrative around AI is shifting from software magic to physical extraction. This issue is crossing party lines. While politicians like Donald Trump call data center concerns a hoax, local Republicans and Democrats are uniting to block construction. As AI companies face a massive trust deficit, developers will need to prove their tools actually deliver real-world value to offset the heavy physical footprint they require.
For builders
Prepare for higher compute costs
Local moratoriums restrict the supply of new data centers. Restricted supply means cloud providers will pass their rising infrastructure costs down to you. Lock in your compute rates now if you rely on heavy model training.
Optimize models for efficiency
The era of throwing endless compute at a problem is facing physical limits. Founders who build smaller, highly optimized models will win. You pay less for infrastructure, and you insulate your startup from hardware shortages.
Monitor state-level regulations
States are taking action while the federal government lags. Track the 35 ballot measures this November. If you build infrastructure-heavy tools, you need to know which states are open for business and which are hostile.
My take
We treat AI like it lives in the ether. It does not. It lives in massive warehouses that suck up local water and power. If the industry keeps steamrolling small towns with secret deals and empty promises, this massive boom will grind to a halt. You cannot build the future if the neighbors refuse to zone it.
Original reporting: Traders Union (FT syndication). This is my rewrite and opinion.