West Texas now hosts two of the most water-hungry industries in the United States at the same time: oil and gas extraction, and the artificial intelligence buildout. Researchers at The University of Texas at Austin argue that the vast volumes of salty wastewater pulled up alongside Permian Basin crude could be treated and redirected to cool the AI data centers multiplying across the region, converting an environmental liability into a usable resource. Accounts of the work cite the journal Nature Sustainability, while a related report places the study in Water Research.
The pitch is deceptively simple: instead of draining freshwater aquifers already under strain, draw on a byproduct the oilfields generate in staggering quantities.
- Enormous supply: the Permian yields roughly 3 to 10 barrels of water per barrel of oil — about 8 billion gallons of “produced water” per day.
- Saltier than the sea: dissolved solids run 100 to 300 grams per litre, against roughly 35 g/L for seawater, forcing high-pressure reverse osmosis or thermal distillation.
- Built-in offsets: waste heat from the data centers and lithium recovered from the brine could help pay for treatment.
- Open risks: induced seismicity and the unresolved question of who funds treatment and transport infrastructure.
Two insatiable demands, one basin
The numbers on both sides are hard to ignore. According to the Environmental and Energy Study Institute, a large data center can consume up to 5 million gallons of water per day, and the Lawrence Berkeley National Laboratory expects data centers to burn through roughly 33 billion gallons by 2028 — the equivalent of nearly 50,000 Olympic-sized swimming pools. Less than 4 percent of Earth’s water is freshwater, and only about 0.5 percent of it is accessible and safe for human consumption.
Set that against the Permian, which generates about 8 billion gallons of produced water daily, a volume more than a thousand times greater than a single facility’s needs. Today that brine — saturated with salts, minerals and traces of hydrocarbons — is hauled long distances and injected underground because no profitable use exists for it.
“The Permian (Basin) puts out 1,200 Olympic-sized swimming pools per day, and right now this water is transported over long distances and injected underground. This is problematic in many ways including the risk of seismic activity,” said Vaibhav Bahadur, a professor in the Cockrell School of Engineering’s Walker Department of Mechanical Engineering and one of the leaders of the research. “We’re also running out of space to bury that water, by as early as 2030. We need to find a way to use it, and cleaning it up for data centers is a great way to do that.”
The treatment problem — and the lithium kicker
Desalination is not new, but Permian produced water is saltier than seawater, which narrows the options. Reverse osmosis is the more energy-efficient route, yet membranes struggle once salinity climbs into the hypersaline range. Heating water into vapour and condensing it back works on almost any feedstock, but carries a heavier energy penalty.
Two features reshape the arithmetic. The first is industrial symbiosis: the waste heat generated by the data centers could partially drive treatment, echoing designs explored in desalination projects elsewhere. The second is lithium. The Permian’s produced water sits atop one of the largest in-solution lithium deposits in the United States, meaning treatment could yield a saleable byproduct rather than pure cost.
“We’re not waiting for a futuristic technology to come along and solve this problem for us,” Bahadur said, noting that the necessary desalination methods already exist and that their expense would represent only a small fraction of a data center’s total energy footprint.
Earthquakes, disposal limits and who pays
The research flags a risk that proponents of the model tend to skip: induced seismicity. Re-injecting produced water has already been linked to rising earthquake activity in Oklahoma and Texas. Diverting those flows toward treatment plants changes the problem but does not erase it if treated water or residual brine is still pumped underground.
The geographical overlap is no accident. The same grid that powers oil extraction is feeding the new campuses, with companies including CoreWeave, Applied Digital and Amazon building large facilities in West Texas.
The decisive question may be political rather than technical. If oil producers absorb the cost of treatment and transport as part of managing their waste, the model looks viable. If data center operators must shoulder it, the bill could exceed simply connecting to a municipal water network in another state. Lithium revenues could tip the balance — provided prices hold near current levels.
The proposal does not claim that oilfield wastewater is a definitive answer to the AI industry’s water problem. It argues something narrower and more useful: that a stream now treated as a disposal burden could become an asset if the right treatment and transport infrastructure is built. For a state where freshwater scarcity and AI capital expenditure are both accelerating, that reframing may prove as valuable as any single technology.
Sources & References