About two-thirds of planned US artificial intelligence datacenters are set to be built in parts of the country that ranked among the driest over the past year, according to a Guardian analysis published Sunday. The finding lands as much of the US remains gripped by a severe drought and as opposition grows to the water demands of the computing facilities needed to train and run AI systems. The projects are spread across drought-hit areas where cooling needs are high and local water stress is already acute. That is the story. And it lands at exactly the point when AI capital spending is accelerating.

The most immediate consequence is political. Water use — not just power demand — is becoming the hard constraint on AI expansion, and communities in dry regions now have a sharper basis to challenge permits, tax deals and project timelines. That matters for investors tracking the same spending boom that has already fueled moves across technology and regional markets, from chip stocks in Korea to financing bets tied to Asian growth and rates. The result: datacenter geography is now a balance-sheet issue, not a side debate.

Background

The core fact is blunt. Datacenters typically need large volumes of water to operate, and the next buildout is heading into dry country anyway. The Guardian said roughly two-thirds of upcoming facilities would be located in places that have been among the driest in the US over the last 12 months. That aligns the industry’s biggest expansion cycle with the nation’s harshest water stress. It also puts local governments in the middle, forced to weigh jobs and tax revenue against household supply, farm use and long-term resilience.

This pressure has been building for months. AI demand exploded first. Then the infrastructure race followed. Companies and developers have hunted for land, grid access and local incentives, often in the same fast-growing interior and Sun Belt regions that already face heat and water strain. Markets have focused heavily on electricity, transmission and chip supply. They should have been looking just as hard at water. Cooling remains essential for many facilities, and dry conditions turn that engineering choice into a public fight.

The backdrop is larger than one industry. The US has been battered by a record-shattering drought, according to the report, and that broad environmental stress is measurable through federal and scientific tracking. The U.S. Drought Monitor and the National Oceanic and Atmospheric Administration have long tracked drought intensity and hydrological strain across states. Datacenters sit inside that reality, not outside it. And as AI becomes more power-hungry, the sector’s water footprint gets harder to dismiss. Even basic cooling system choices now carry political and financial consequences.

What this means

The next stage is straightforward. More projects will face tougher local review, more disclosure demands and louder resistance from residents who don’t want scarce water redirected to server farms. That is where the economics change. A datacenter approved quickly in a water-secure region may now be worth more than a larger site stuck in a county fight over aquifer draw and summer restrictions. Investors chasing AI infrastructure growth have treated land pipelines as inventory. They’re actually regulatory options with hydrology attached.

But this cuts both ways. Dry-state officials still want the investment, and developers know it. So the likely response is not a broad retreat from drought-hit markets. It is a scramble to redesign projects around lower-water cooling, recycled water arrangements, tougher reporting and narrower promises on withdrawals. Some plans will survive under those terms. Others won’t. The winners will be operators that can prove they need less water per unit of computing output. The losers will be projects built on the old assumption that communities would swallow the trade-off because AI sounds strategic.

There is a broader precedent here. Once water becomes a headline risk for AI, every permit battle gets easier for opponents and more expensive for sponsors. The same buildout that has reshaped bond markets and corporate funding plans — visible in moves from Taiwan bond yields to central-bank support measures such as India’s FX swap window — now has a local resource problem attached to it. This is no longer an abstract sustainability argument. It is a siting problem with legal teeth. And it will force the industry to choose between faster deployment and lower water intensity.

Water use is becoming the hard constraint on AI expansion.

Key Facts

  • About two-thirds of planned US AI datacenters are slated for areas among the driest in the country over the past year, according to the Guardian analysis published June 8, 2026.
  • The projects involve datacenters, facilities that typically require large volumes of water to operate and cool computing equipment.
  • The report says the buildout is advancing even as a record-shattering drought has affected much of the United States.
  • The dispute centers on water demand tied to AI growth, adding to existing scrutiny of energy use, grid strain and local permitting.
  • Federal drought conditions are tracked by agencies and research partners including Drought.gov, while wider climate and water science is covered by sources such as the EPA and the data center reference archive.

What comes next is specific. Watch local permitting fights, water board reviews and county-level tax incentive debates around proposed sites over the next several months, because that is where this national AI spending story now gets priced. If opponents can tie new facilities to measurable drought stress, approvals will slow and project costs will rise. And if developers answer with stricter water plans, that will become the new baseline for every datacenter proposal filed after June 2026. (The companies involved have not responded to every public concern raised in recent months.)