Product1 distinct publisher3 min readUpdated
Liquid Death's ad with Jason Kelce is a gag, but substituting treated non-potable water in data center cooling is routine. The binding constraint is not water rights; it is a big enough treatment plant nearby.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
Liquid Death and Garage Beer released a campaign in which former Philadelphia Eagles player Jason Kelce asks the public to donate urine to cool AI data centers, opening with the line that "AI data centers waste millions of gallons of water" [1][2]. The gag points at something operators are already doing in a much duller form: running cooling systems on treated non-potable water instead of drinking water [3][4].
Michael Obradovitch, vice president of Data Center Global Accounts at Ecolab, told TechCrunch that a fair amount of alternative water is already being used to cool these facilities [3]. Those sources at least partially offset demand on potable supply [4]. Recycled water is wastewater and sewage treated until it is safe to reuse, and wastewater does contain human urine, which is the thread the ad accidentally pulls [5]. Bruno Pigott, executive director of the WateReuse Association and a former acting assistant administrator for water at the EPA, put the distinction plainly: you would not use urine as-is, but you can clean it into useful water [6].
The direct version fails on chemistry. Pigott said urine carries salts, urea, bacteria and organic matter that leave mineral deposits, and feeding that into a cooling tower would demand constant cleaning [7]. Evaporative cooling works by passing hot air through water to strip heat [8]. Urine can be turned into potable water, which is what astronauts do, but that is not efficient at scale [14]. Municipal plants do the work instead, using membrane bioreactors, reverse osmosis, ultraviolet light and other steps to reach industrial quality, and sometimes drinking quality [9].
This is not new technology; it is newly contested capacity. Greta Zornes, water reuse practice leader at CDM Smith, said recycled water has been used for industrial cooling for decades, and that there has been a boom in it for data centers [10]. Zornes has worked in water reuse for more than two decades and said she now works on recycled water for data centers every day [11].
Which is where siting arguments are being framed wrong. The negotiation is usually conducted as a volume dispute - how many gallons the campus will draw. The more decisive variable is adjacency. Zornes said a site has to be reasonably near a wastewater treatment plant large enough to supply it, and that rural plants often are not treating enough water to serve a data center at all [12]. That means the same nameplate load is a different water proposition in Loudoun County, Virginia, which hosts more than 250 data centers and has plans for at least two dozen more, than on a greenfield site an hour from any plant of size [13][15]. A county without reuse infrastructure is, by default, negotiating over its drinking water.
One caution on the numbers: none of the practitioners quoted put a figure on how much data center cooling water is recycled [16]. "A fair amount" and "a boom" are not volumes. An operator or a county staffer reading a water commitment should ask for gallons per day of non-potable supply, the named source plant, that plant's current treated throughput, and what happens to the substitution rate during a drought or an outage at that plant.
What to watch: whether new development agreements specify a non-potable share rather than a total draw, and whether the push toward rural sites quietly shifts load back onto potable systems [12]. Loudoun's next two dozen facilities are the cleanest test of whether reuse capacity scales with the buildout [13].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Michael Obradovitch, vice president of Data Center Global Accounts at Ecolab, told TechCrunch that while the Liquid Death commercial is tongue-in-cheek, in reality there is a fair amount of alternative water sources already being used to a similar extent to cool these data centers.
Alternative water sources used in data centers at least partially offset the demand for potable drinking water.
Zornes said a facility has to be somewhat near a wastewater treatment facility sizable enough to supply enough water, and that when data centers go into rural areas the wastewater treatment plants often are not big enough or treating enough water to be used.
Liquid Death teamed up with Garage Beer and former Philadelphia Eagles star Jason Kelce on a marketing campaign proposing that people's urine be used to cool AI data centers.
In the video campaign, Kelce says: "AI data centers waste millions of gallons of water."
Recycled water is made by treating wastewater and sewage water so they are safe to use again, and wastewater and sewage contain human urine.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Named practitioners plus one utility dataset, single outlet
Three on-the-record domain experts with directly relevant roles (Ecolab data center accounts, WateReuse executive director and former acting EPA water administrator, CDM Smith water reuse practice leader) support the mechanism and practice claims, and one named utility (Loudoun Water) supplies the only numbers. But everything arrives through a single publisher, no primary documents, permits or utility filings are linked, and the practitioner testimony is qualitative — no share or volume of recycled cooling water is given at sector level.
Real and scaling, but majority-potable in the one measured market
Recycled water for industrial cooling is described as decades-established, a practitioner reports a current boom specifically in data center reuse, and Loudoun County demonstrates deployment at scale — roughly 200 million gallons per day of recycled water. Adoption is bounded by the same evidence: 57% of that county's data center water is still potable, and rural siting frequently lacks a treatment plant large enough to reuse from.
Modestly overstated routineness
The advertising premise is openly a gag and the reporting corrects it rather than amplifying it, which keeps the gap small. The residual overstatement is in the framing of non-potable substitution as routine: the only quantified market still draws the majority of its data center cooling water from potable supply, and no practitioner offers a figure for how widespread reuse actually is, so 'already normal' runs somewhat ahead of what the evidence shows.
Every quoted voice sells or lobbies for the solution
The sourcing is uniformly interested. Ecolab sells water treatment services into data centers, CDM Smith is an engineering firm whose practice leader says she works on data center recycled water daily, and WateReuse is an advocacy body actively lobbying for a 30% tax credit for industrial recycled-water infrastructure. The originating campaign is beverage-brand marketing, and Meta's cited $270 million wastewater commitment is a corporate community-relations disclosure. No independent regulator, ratepayer, or critical hydrologist voice appears to offset these interests.
Directionally solid, thinly corroborated
The mechanism and constraint claims are internally consistent, specific, and come from identifiable experts, and the story is fresh. Confidence is held down by single-publisher sourcing, uniformly interested informants, one geographic data point standing in for the sector, and no primary documentation of the Meta investment or the tax-credit proposal.
product
Pew's under-30 numbers turn AI hostility into a positioning constraint1 distinct publisher
product
The dirtiest part of the AI gas buildout is a turbine spec, not a nameplate1 distinct publisher
invest
Data-center politics graduated from zoning boards to primary ballots, and capex plans have not repriced1 distinct publisher
product
Nebius funds $4.5bn of AI capacity on terms that pay lenders mostly in stock2 distinct publishers
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 20, 2026