Water
The verifiable public record on Google data center water use, what it does and does not tell us about Three Way Road, and the questions still unanswered.
What "air-cooled" has meant elsewhere
Google publishes water consumption by site but does not label which sites are air-cooled, so no true air-cooled average can be calculated from public data. Reporting has identified two.
That is the most important fact on this page. Two facilities, the same cooling category, a 150-fold difference. "Air-cooled" by itself tells you almost nothing about how much water a site will use. Size, staffing, climate, and what the operator counts all move the number more than the cooling label does.
Secondary Both site figures come from an Associated Press–originated analysis of company sustainability disclosures, widely republished in 2025. Google publishes site-level water data but does not label which sites are air-cooled. The air-cooled attribution is the analysts', not Google's. Verify against Google's own environmental report before quoting these at a podium.
The trend
Google withdrew 7.8 billion gallons in 2024 and consumed 78% of it. Of 24 disclosed sites in the U.S. and Canada, nine withdrew under 500,000 gallons a day and three withdrew more than 2 million. The largest, Council Bluffs, Iowa, withdrew an average of 3.9 million gallons a day and consumed 2.8 million.
Secondary Fleet totals and the site distribution are compiled from Google's 2025 and 2026 environmental reports by third-party analyses. The underlying reports are Primary and worth reading directly.
Replenishment is a real commitment and Google has funded it. It is also not the same thing as leaving water in the Wichita River basin. Replenishment projects can sit in other watersheds entirely.
Two precedents worth knowing
A Virginia newspaper sued the local water authority for records on a planned Google data center campus. Those records showed officials had agreed to supply the project with up to 2 million gallons of water daily by January 2028, with the potential to rise to 8 million gallons daily depending on future expansion.
Reported Reported from records obtained by a Virginia newspaper through litigation against the local water authority.
This is the closest analogue to our situation: a planned Google campus, where the contracted volume only became public through litigation. It is the reason we are asking to see the WVWSC agreement now rather than later.
After a newspaper's lawsuit forced disclosure, records showed Google's water consumption in The Dalles, Oregon quintupled between 2012 and 2025, reaching about 550 million gallons, nearly 40% of the entire city's water consumption last year.
Reported Reported from documents disclosed after an Oregon newspaper's lawsuit, which the parties settled with an agreement to disclose annual water use.
The lesson is not that the same will happen here. It is that a data center campus is not a fixed thing. It phases, it expands, and the first-year number is not the number you live with.
The question we are asking
Google's Project Raptor FAQ gives a maximum daily water use of 32,000 gallons, described as domestic use (restrooms, kitchens, facility maintenance) for a campus with 50 permanent employees.
Primary The 32,000 gallon figure, the domestic-use description, the 50 permanent jobs, and the unquantified one-time closed-loop fill are all from the Project Raptor FAQ.
Secondary The annual projection and per-employee figure are our own arithmetic on those numbers, shown so you can check it: 32,000 × 365 = 11.68 million; 32,000 ÷ 50 = 640. Typical office water use of 15–25 gallons per person per day is a widely published planning figure, not a measurement at this site.
We want to be fair about this. A maximum-day figure is an engineering number used to size pipe, not a prediction of average use, and conservative sizing is exactly what you want in a water system. Raptor is also a far larger campus than either site above.
But 640 gallons per person per day is roughly thirty times a normal office, which means the 32,000 is either very conservative or it covers more than the FAQ itemizes. Two things we know it does not cover: the one-time fill of the closed-loop cooling system, which the FAQ calls relatively large without giving a number, and construction water for concrete mixing and dust control.
So the ask is simple and answerable: what is the projected annual average water use, what is the one-time closed-loop fill volume in gallons, and what is total construction water? None of those are unreasonable questions and none of them require anyone to be against the project.
The bigger number
Lawrence Berkeley National Laboratory estimated that U.S. data centers consumed 17.4 billion gallons directly through cooling in 2023, and an additional 211 billion gallons indirectly, through the water consumed generating their electricity. Roughly twelve times more. Company disclosures do not consistently include it.
Estimate Lawrence Berkeley National Laboratory, 2024 report to Congress. A modeled national estimate across many facilities, not a measurement of any one site.
For a 1,000 MW load, the indirect footprint almost certainly dwarfs anything measured at the meter on Three Way Road. The Wilbarger County project gets its own co-located generation. The Archer County plan is a direct connection to the Bowman switch station on the grid. Where those electrons come from, and what they cost in water, is a fair question for ERCOT and the PUCT.
Estimate HARC white paper, January 2026. Also a model, covering 464 facilities statewide.
For scale in Texas: a January 2026 HARC white paper estimated the state's data centers consumed about 25 billion gallons in 2025 counting both direct cooling and indirect power-plant use, across 464 existing facilities, roughly 0.4% of statewide water use.
Our standard
Air cooling is a genuine improvement over evaporative cooling on water, and Google choosing it here is not nothing. The disclosed air-cooled sites use a thousandth of what the big evaporative sites use. We are not going to pretend otherwise.
We are also not going to repeat the industry rules of thumb circulating online that would imply far higher figures for a campus this size. We found several. None of them published a methodology, so none of them are on this page.
If we put a number in front of you, you can open the document it came from. That is the whole standard, and it is why the questions above are worth answering.