The race to dominate AI infrastructure just got a new subsidy: your lungs.
The Summary
- Trump's EPA is dismantling environmental regulations to accelerate AI data center construction, prioritizing speed over public health
- Former EPA officials warn of increased pollution risks and are pushing a "Data Center Health Protection Pledge" — likely to be ignored
- EPA Administrator Lee Zeldin framed deregulation as making America "the AI capital of the world"
The Signal
The computing power required to train and run frontier AI models is staggering. A single GPT-4 training run consumed enough electricity to power 1,000 homes for a year. That power has to come from somewhere, and right now "somewhere" means fossil fuel plants running hotter and data centers burning through water reserves for cooling.
Trump's EPA just decided that speed matters more than the air you breathe. Administrator Lee Zeldin has rolled back dozens of environmental rules since taking office, specifically to remove barriers to data center construction. The logic is simple: if America wants to lead in AI, it needs more compute capacity, faster. Environmental reviews slow things down. So cut them.
"Making America the AI capital of the world through deregulation means someone pays the externality cost."
Former EPA officials are sounding alarms. They released a report this week detailing health risks from accelerated data center construction: increased particulate matter from power generation, water table depletion in drought-prone regions, and thermal pollution from cooling systems. Their proposed "Data Center Health Protection Pledge" asks the administration to at least monitor and disclose environmental impacts. The White House has not responded.
This is Web4 infrastructure policy in its rawest form. The agents need compute. Compute needs power. Power needs permits. So we're changing the permits. What's interesting is not that this is happening, it's how explicitly the trade-off is being framed. Zeldin is not pretending there is no cost. He is saying the cost is worth paying to win the AI race.
The key tensions here:
- China is building data centers without environmental constraints. America either matches that speed or cedes AI dominance.
- Local communities bear the pollution cost while tech companies and their shareholders capture the value.
- The compute infrastructure being built now will determine who controls AI agents for the next decade.
The irony is thick. Web3 promised decentralization to escape corporate control. Web4 is being built on centralized data centers that require regulatory capture to scale fast enough. The "own" part of Read+Write+Own+Build assumes you can breathe while you build.
The Implication
If you are building on AI infrastructure, understand that your speed advantage comes with a subsidy. That subsidy is currently being paid by people who live near data centers and by future generations who will inherit depleted aquifers and dirtier air. Eventually, that bill comes due, either through new regulations or through infrastructure bottlenecks when water runs out.
For investors, this signals which regions will see the fastest data center growth: places with weak state-level environmental rules and desperate local economies willing to trade clean air for jobs. For residents near proposed data center sites, organize now or get steamrolled. For everyone else, watch what happens when "move fast and break things" meets physical infrastructure at scale.