If you've seen the coverage, the concern is understandable: a single large data center can use millions of gallons of water a day, cooling racks of servers that never stop running. Local news has covered specific towns fighting specific projects over exactly this. It's not a manufactured panic. It's a real number attached to a new, unfamiliar-feeling neighbor.
But a number only means something next to other numbers. And nobody putting "millions of gallons a day" in a headline is putting it next to what else uses water in this country.
The comparison nobody puts next to it
All US data centers combined used about 17.4 billion gallons of water directly in 2023, according to EPA-backed research cited by industry analysts. That sounds like a lot until you look at what's next to it.
Sod farms — the commercial operations that grow grass specifically to sell as turf — cover 339,551 acres in the US according to the 2017 USDA Census of Agriculture. At a typical irrigation rate of around 600,000 gallons per acre per year, that's an estimated 204 billion gallons a year spent growing grass before it's ever installed on anyone's lawn — more than ten times what every data center in the country uses directly.
Golf courses use more still: roughly 425 to 547 billion gallons a year, by some counts thirty times the data center figure.
California's almond orchards, on their own, use about 1.3 trillion gallons a year — roughly as much as every household in the state combined.
Zoom out to lawn and landscape irrigation nationwide — mostly residential — and the EPA estimates close to 9 billion gallons a day, or about 3.3 trillion gallons a year.
And all of that is still a rounding error next to agriculture as a whole. Irrigation accounts for 42 to 47 percent of all US freshwater withdrawal, out of a national total of roughly 322 billion gallons withdrawn per day — around 117 trillion gallons a year. That puts agricultural irrigation somewhere in the neighborhood of 49 to 55 trillion gallons annually.
| Water user | Annual US usage (gallons) |
|---|---|
| All US data centers (direct, 2023) | ~17.4 billion |
| Sod farms, grown for sale | ~204 billion (estimated) |
| Golf courses | ~425–547 billion |
| California almonds alone | ~1.3 trillion |
| Lawn & landscape irrigation (nationwide) | ~3.3 trillion |
| All agricultural irrigation (nationwide) | ~49–55 trillion |
| Total US freshwater withdrawal | ~117 trillion |
It takes roughly 1.1 gallons of water to grow a single almond. California's almond orchards alone use about as much water each year as every household in the state combined.
Data centers are a small fraction of a fraction of a fraction. If water conservation were really the driving concern, the attention is landing on one of the smallest targets in the room — smaller than sod, smaller than a golf course, smaller than a snack food.
Why data centers get the attention anyway
This isn't cynicism, just an honest read of why the outrage lands where it does. Data centers are new. They're visible. They're wrapped up in AI and big tech, which already carries its own set of anxieties. And they tend to arrive as single, large, controversial projects that make local news — a specific building, a specific town, a specific fight.
Golf courses, sod farms, and almond orchards are old, familiar, and diffuse. Nobody's town hall gets packed over the twelfth green's irrigation schedule. The attention gap tracks with visibility and novelty, not with actual water impact.
This is also a temporary problem
The 17.4 billion gallon figure isn't a steady state. It's closer to a peak, on its way down, for reasons that have almost nothing to do with the water debate.
The AI chips driving the current data center boom run far hotter than anything air cooling was designed for — 30 to 60 kilowatts per rack, well above the 10 to 15 kilowatt ceiling where air cooling stops working. That's forcing a shift to closed-loop liquid cooling, which circulates the same water in a sealed system instead of evaporating it into the air. The water difference is not incremental: a comparable-scale campus using closed-loop cooling uses about 22,000 gallons a day, versus roughly 5,000,000 gallons a day for an evaporative campus of the same size — more than a 200-fold drop.
Microsoft's newest AI campuses, starting with Atlanta (live since October 2025) and Wisconsin (coming online in early 2026), are built on closed-loop cooling from the ground up. Google, Meta, and Amazon are rolling out liquid-cooled facilities the same way. Nobody is doing this primarily as a water-conservation gesture — they're doing it because the chips leave no other option. The water savings are a side effect of a physics problem, not the goal of a PR campaign.
The incentives lining up around this are, so far, mostly getting the shape right. A federal bill would offer a 30 percent tax credit for water reuse investments, technology-neutral, available to any operator regardless of size. California's proposed data center tax credit rewards measurable outcomes — carbon-free power, water-efficient cooling — rather than mandating a specific method. A handful of states are moving to require water-use disclosure rather than dictate a technology. That combination — reward the outcome, disclose the impact, let operators pick the method — pushes the whole industry toward less water use without locking out smaller operators who can't yet absorb the 20 to 40 percent higher upfront cost of a full closed-loop build. A couple of states are considering outright mandates instead, which risk the opposite: freezing in place whichever companies can already afford the premium.
Either way, the number at the top of this post is shrinking, not growing, for the parts of the industry building anything new.
What blocking it actually accomplishes
Here's the part that matters more than the comparison. Local opposition has already blocked or delayed roughly $64 billion in US data center projects since 2024. That demand doesn't evaporate. It moves — and it's already moving, right now, to Querétaro, Mexico.
- Since 2020, over $12 billion has been invested in Querétaro data centers — from Microsoft, Google, and Amazon, the same companies building, and getting blocked, in the US
- The region is enduring its worst drought in 100 years
- Eight of Querétaro's eleven regional aquifers are already overexploited
- Some towns near the data centers are rationing water to three days a week
- The state government exempts facilities built in industrial parks from standard environmental impact reporting
Block the data center in Virginia, and the demand doesn't vanish — it moves to a drought-stricken Mexican state with weaker reporting requirements, built by the same company.
This isn't a hypothetical about some other country doing it worse someday. It's happening in parallel, right now, with the same corporate names on the buildings.
The honest version of the argument
This pattern isn't universal. Some countries and regions have real environmental standards for data center construction; some don't; it varies even within the same country. Blocking a project in the US doesn't automatically send it somewhere worse in every case.
But the general dynamic holds regardless of the specific destination: global AI compute demand isn't something a local zoning board can switch off. A US county can decide whether the growth happens under US environmental review, US labor law, and US utility oversight — or somewhere else, without any of that. Querétaro is simply the clearest live example of what "somewhere else" looks like right now.
The actual tradeoff
If the goal is genuinely conserving water, the attention is misallocated. Lawn irrigation and sod farming alone dwarf what data centers use, and neither gets a fraction of the scrutiny.
If the goal is reducing the environmental footprint of AI, blocking construction in your town doesn't shrink that footprint. It relocates it — to a place with less regard for getting it right, under a government less inclined to ask questions, while the jobs, tax revenue, and economic upside move with it.