The cloud isn't as clean as the name suggests, and Ohio is about to make that literal.
The Ohio EPA is weighing a proposal that would allow data centers to discharge wastewater directly into the state's rivers. The timing isn't coincidental—Ohio has become a hotbed for data center construction, driven by cheap electricity and favorable tax treatment. But those massive server farms generate enormous amounts of heat, which means enormous amounts of water needed for cooling.
Here's what's in that wastewater: chemicals used to prevent bacterial growth in cooling systems, dissolved metals from pipes and equipment, and thermal pollution that can raise river temperatures enough to affect aquatic ecosystems. The proposal would set discharge limits, but environmental groups argue those limits are based on outdated assumptions about how much dilution occurs in rivers, especially during low-flow periods.
The AI boom is making this worse. Training large language models requires computational resources that translate directly into cooling requirements. A single data center can use millions of gallons of water per day. Multiply that by the dozens of facilities being built or expanded across Ohio, and you're talking about water usage that rivals small cities—except cities treat their wastewater before releasing it.
Data center operators argue they're following all applicable regulations and that the discharged water meets safety standards. That's technically true, but it's also true that those standards were written for a different era of industrial activity. The question isn't whether data centers are breaking current rules—it's whether the rules are adequate for the scale of infrastructure we're building.
There's a deeper irony here: we're funding the AI revolution by potentially degrading the physical environment those AIs will eventually optimize. The technology is real, the efficiency gains are real, but the environmental costs are real too. Ohio is about to decide whether rivers are an acceptable place to externalize those costs.
