Training and operating large AI models requires immense computing power—and keeping those servers from overheating demands vast quantities of water. Tech giants like Google, Microsoft, and Amazon are building or expanding data centers across the globe, often in water-stressed regions. The result? AI is becoming not only an energy-intensive industry but also a surprisingly water-hungry one.
A 2023 study by researchers at the University of California, Riverside, estimated that training GPT-3, a single large language model, consumed 700,000 liters of clean freshwater—enough to cool a nuclear reactor for several hours or produce hundreds of Teslas. And that figure does not include the ongoing costs of running the model once deployed.
Microsoft reported that its global water use jumped by 34% in 2022, largely due to its AI ambitions. Google, too, saw its water consumption climb by 20% in the same period. While both companies claim to be investing in water stewardship programs and pursuing “water-positive” goals, the data paints a different picture: AI is accelerating demand for cooling infrastructure that draws from rivers, lakes, and municipal supplies.
The problem is especially acute in places like Arizona, where Microsoft has built major data centers amid a historic drought. Locals and environmental advocates are raising concerns about transparency, accountability, and long-term sustainability.
Unlike energy consumption, which has become a central focus of climate-conscious innovation, water usage in AI has remained relatively underreported. There are no standardized disclosure rules, and few consumers realize the environmental footprint of a single AI query—which, in some cases, can be dozens of times higher than a Google search.
As investors grow increasingly focused on ESG (environmental, social, and governance) metrics, this hidden cost may soon come under closer scrutiny. Regulation could follow, particularly in Europe, where water use is more tightly monitored and companies face stronger environmental compliance requirements.
The race to dominate AI is heating up, but without careful resource planning, it may also leave scorched earth behind. Water, long overlooked in the digital economy, could become the next battleground in tech’s relentless push for scale.