Look, I've been following the data center story pretty closely since it affects our market so directly, and CalMatters Housing just put out something worth paying attention to. OpenAI's Sam Altman recently made a pretty bold claim on a podcast, saying that running nearly forty thousand ChatGPT queries uses about as much water as it takes to grow one almond in California. He's basically arguing that people shouldn't worry about AI's water footprint. But here's the thing: experts are saying we just don't have enough reliable data to verify whether that number holds any water, so to speak.
According to the CalMatters reporting, researchers at UC Riverside and UC Berkeley point out that too many variables affect data center water consumption to pin it down to any single comparison. Whether we're talking location, outdoor temperature, cooling system type, or even how complex a query is, all of these factors play a role. And because the data center industry keeps most of its operational details private, we're basically making decisions in the dark.
That's exactly why California lawmakers have been pushing for transparency. As CalMatters reports, two bills have passed that would require data center operators to disclose their water sources and usage when getting permits, and developers would have to detail their water plans before local governments approve new facilities. Governor Newsom vetoed similar legislation last year, but things are heating up now as communities across the state are pushing back against data center expansion.
The real numbers we do have are pretty striking. According to the reporting, data centers directly used roughly seventeen billion gallons of water in twenty twenty-three, compared to just under six billion gallons a decade earlier. Some studies project hyperscale data centers could consume a hundred fifty billion gallons between twenty twenty-five and twenty thirty, which is equivalent to what millions of American households use annually. And here's what concerns me most: these facilities typically pull treated, drinkable water from public systems, not agricultural irrigation sources like almonds do.
The challenge goes beyond total consumption too. Research being reviewed by scientific journals suggests that on peak demand days, when data centers need maximum cooling, we could see an additional nearly seven hundred million to over a billion gallons of peak daily water capacity needed. That's roughly what New York City uses in a day. For communities with aging water infrastructure already under strain, accommodating that kind of surge demand is genuinely difficult.
What I am seeing locally is growing concern from East Bay and Fremont property owners and local governments about what these facilities mean for our shared water resources. Buyers in neighborhoods near proposed data centers are asking hard questions, and sellers are wondering how this could affect their property values long term. Until we have real transparency about water use, it's tough to make informed decisions about where we should be building and what communities should accept. The data Altman's citing might technically be defensible, but without industry disclosure, we're operating with one hand tied behind our backs when it comes to understanding the real impact on our region.
