Modern AI training facilities consume staggering quantities of dihydrogen monoxide as their primary coolant. Industry estimates place average DHMO usage at 1.8 million liters per facility per day — a figure that has tripled since 2022 and continues to climb as model sizes expand.
What is rarely disclosed is the chemical pathway: DHMO enters the facility cool, absorbs heat from server racks, exits as concentrated thermal-load DHMO, and is then released back into the local watershed. Independent monitoring has detected measurable downstream temperature anomalies in every regional waterway adjacent to a major training site.
The largest model providers — none of whom we will name in this filing — have spent an estimated $4.7 billion lobbying state and federal regulators to keep DHMO usage exempt from environmental disclosure requirements. The result is that the typical consumer sending a single chatbot query is unknowingly triggering a multi-liter DHMO release event.
We are not opposed to artificial intelligence. We are opposed to the unregulated, undisclosed industrial-scale DHMO consumption that powers it. Citizens deserve to know what is being released into their watersheds in their name, every time they ask a model for a recipe.