'Data Centres Use Less Water Than Golf Courses.' True Today — and About to Flip.

Kevin O'Leary's stat checks out: US golf uses 2.08B gallons/day, data centres ~449M — 4.6x more. But golf's curve is flat and data centres are exponential; the lines cross around 2028. Using a true present-tense fact to dismiss a problem defined by its growth rate is the move to watch. His own Utah project shrank 75% under pressure.

A technically-true deflection: a real number that makes your footprint look small next to something frivolous.

Kevin O'Leary has a comeback for anyone worried about AI data centres draining the water supply: golf courses use far more. He's right — today. He's also using a true number to wave away a problem that the same numbers say is about to become his.

The claim, and the honest scorecard

The Shark Tank investor's argument is that American data centres consume far less water than the country's golf courses. The figures back him up, for now, per The Next Web: US golf courses use about 2.08 billion gallons a day; US data centres, roughly 449 million. Golf uses about 4.6 times more. As a factual snapshot, O'Leary is correct.

The problem is that a snapshot is exactly the wrong tool for this question. Water stress isn't about who uses more today — it's about the trajectory, and the two lines are heading in opposite directions.

Why the comparison misleads

Golf's water consumption is flat to declining. Courses are switching to drought-resistant grasses and recycled water; the sport has every incentive to shrink its footprint and is doing so. It's a mature use case whose demand curve bends downward.

Data-centre water demand is on an exponential curve, driven by AI training and inference. And the projections cross. Data-centre consumption is expected to reach roughly 590 billion gallons by 2028, potentially surpassing golf's projected ~425 billion around 2026–2027. So the honest version of O'Leary's claim is: "Data centres use less water than golf courses — for about another year." Using a true present-tense fact to dismiss a problem defined entirely by its growth rate is the rhetorical move to watch for. The gap he's citing is closing as he cites it.

There's also a distribution problem the national totals hide. Golf courses are spread across the country; data centres cluster, and increasingly in already water-stressed regions. A billion gallons averaged nationally is a very different thing from a concentrated draw on one strained local aquifer. The aggregate comparison flatters the data centre precisely by averaging away where the strain actually lands.

The tell is in his own project

The most instructive part of the story is O'Leary's own behaviour. His 40,000-acre Stratos data-centre project in Utah sparked protests and a gubernatorial executive order, and he scaled it back by 75% — to 10,000 acres — after pressure from Republican state senators. If water use were the non-issue his golf comparison implies, a 75% reduction under political pressure is a strange thing to concede.

He says the facility will use a closed-loop chilling system with no continuous water draw — which, if true, would genuinely address the concern. But Virginia Tech experts say there isn't enough data to verify the claim. That's the crux: "closed-loop, no water draw" is the right answer to the water problem, and also an easy thing to assert and a hard thing to prove. The engineering solution exists; whether a given facility actually implements it as advertised is the question that transparency, not reassurance, resolves.

What the argument is really doing

The golf comparison belongs to a familiar genre: the technically-true deflection. Find a real number that makes your footprint look small next to something frivolous, and you can dismiss a concern without engaging its substance. It works because the fact checks out and golf is an easy target. It fails because the concern was never "is this the single largest water user today" — it was "is this a fast-growing, geographically concentrated new draw on a stressed resource, and is it being built responsibly." O'Leary answered a question nobody was asking.

The read

Data centres versus golf courses is a genuinely useful comparison — it just proves the opposite of what it's deployed to prove. Yes, golf uses more water today. No, that doesn't make data-centre water use a non-issue, because golf's curve is flat and the data-centre curve is exponential and they cross within a couple of years. The productive conversation isn't "who uses more" — it's closed-loop cooling, honest local-level disclosure, and siting facilities where the water can actually spare it. O'Leary's own 75%-smaller Utah project suggests he knows the real answer isn't a golf statistic. It's building the thing responsibly and being able to prove it.

Analysis of public figures and projections as reported in July 2026; water-use estimates vary by source and methodology. Not investment advice. Source linked above.

https://redrobot.online/2026/07/19/data-centres-use-less-water-than-golf-courses-true-today-and-about-to-flip/