A Third of What? The Dalles Water Figures Have No Common Denominator

The figure everyone quotes

The Dalles, Oregon, a city of about 16,200 people, hosts three Google data centre sites¹. The claim that travels with it is that Google takes "a third" of the city's water. The claim is roughly right in direction and unreliable in its arithmetic. Three different share figures now circulate, each with a different year, a different numerator and no stated denominator. For anyone who has to plan around local water as a priced and rationed input, the lesson is not about Google. A share figure without a stated base cannot carry a decision. That holds whichever way it points.

What the records show

The firmest numbers come from the city itself. In response to a public-records request, The Dalles gave OPB a spreadsheet on 24 December 2025. According to it, Google used 383.9 million gallons in 2023 and 434.4 million gallons in 2024. That is 1.05 million and 1.19 million gallons a day². On the same record, annual use rose from 104 million gallons in 2012, an increase of 316%². The spreadsheet also lists the city's total consumption each year. OPB published it only as an image, and its totals do not appear in the text².

These are city-held records, not operator claims. The city does not say what they measure. "Used" could mean water delivered to Google's meters or something else, and the published text does not settle it.

Google reports its own figure. Its 2026 Environmental Report says the company "consumed 469 million gallons of water in The Dalles in 2025"⁴. The report does not define "consumed" in the passage that carries the number. Nor does it say whether the figure is withdrawal, delivered supply or evaporative loss. This is an interested party's figure, reported in a document that also presents The Dalles as a site where water cooling is "a suitable, energy-efficient solution"⁴.

Four shares, four bases

Each row is a different measurement. None can be checked against another without the missing denominator.

The arithmetic does not reproduce "a third". Take the city's own 434.4 million gallons for 2024. Divide it by the 4.2 million gallons a day that Shah gives as the city's average demand for that year, which is 1,537 million gallons over 366 days. The result is 28.3%²³. Shah's own numerator for 2024 is 461.1 million gallons, about 6% above the city figure OPB reports, and he puts it at 30%³. Shah calls the same quantity "withdrawals" in his figure caption and "demand" in his text. For 434.4 million gallons to be a third, the city's total would need to be about 1,303 million gallons, or 3.6 million gallons a day. That may be the "consumption" column in the spreadsheet OPB holds. It is not the demand figure Shah cites. On available text, the two bases are about 15% apart.

The 29% has no origin. ScienceBlog's headline-level summary says Google's data centres were "drinking 29 percent of the town's water". Its body says "roughly a third" and never gives 29% or a source for it⁵. The same page also contradicts itself on who sued whom⁵.

The 2025 figures do not agree with each other. Latitude Media reports about 550 million gallons, "nearly 40%" of the city's total consumption, and links to OregonLive coverage of the records⁶. Google's own report gives 469 million gallons for the same year⁴. The gap is about 81 million gallons, or 17%, and no source explains it. They may measure different things, such as delivered supply against evaporative consumption. That is plausible but unconfirmed.

The 2024 share is between about 28% and 33%, depending on which city total is used. The 2025 share cannot be stated until the two 2025 numerators are reconciled.

What the dispute is about

The numbers became contested because they are tied to an expensive plan. The city gets most of its water from the Dog River. OPB first said almost 85%, and later noted that public documents give 78% to 85%¹². The city wants to acquire national forest land around its reservoir and expand storage from 900 to about 3,000 acre-feet, at an expected cost of about $70 million¹. The wider programme of water upgrades is put at about $260 million over 20 years. According to OPB's reading of the city's 2024 master plan, residential water rates are projected to rise 99% by 2036¹. OPB first printed 7% and corrected it¹².

The mayor, Richard Mays, says Google is "really not a factor" in the reservoir plan¹. The retired public works director, Dave Anderson, called OPB's coverage "egregiously misleading". OPB found one error after a line-by-line review, corrected it and stood by the rest². No source here attributes the rate rise to Google. Anderson himself calls Google the city's largest water customer. The 2024 master plan provides for an unnamed "large industrial user" of about 1 million gallons a day, though Anderson says this user is not Google².

So the share figure is doing political work. Both sides are arguing over a ratio whose base neither has published in checkable form.

What this establishes, and what it does not

It establishes three things. Google's use in The Dalles is large and rising on the city's own records. It roughly quadrupled from 2012 to 2024². And on any of the available bases it is at least a quarter of local supply.

It does not establish four things:

• which share figure is correct;
• whether city and Google figures measure the same quantity;
• why the two 2025 figures differ by 17%;
• whether Google's growth is driving the reservoir and rate plans the city says it would pursue anyway.

The control gap is clear. The settlement of the Oregonian/OregonLive public-records lawsuit committed the city and Google to disclosing annual water use⁶. Google now promotes a framework that includes annual disclosure as an industry standard⁷. Disclosure has happened. A common definition has not.

The judgement

Before a share-of-supply figure goes into a siting, permitting or community-risk decision, apply one test. Name the year and the metric: withdrawal, delivered supply or consumption. Then name the city total it is divided by, and who published both. If any of the three is missing, the figure is a talking point, not a planning input.

For The Dalles, the defensible statement today is narrower than the headline. In 2024, Google took 434.4 million gallons on city records, between roughly 28% and a third of city water depending on the base. In 2025, two incompatible figures exist and neither has been reconciled. A disclosure regime that requires the metric and the denominator alongside the number would settle this local argument in one table. Without one, the same argument will repeat at the next site.