What would disprove this analysis? (Criterion 1)

Data center water consumption decreases despite growing compute demand through efficiency improvements

What would disprove this analysis? (Criterion 2)

Data centers in water-stressed regions do not measurably impact local water availability

What would disprove this analysis? (Criterion 3)

Alternative cooling technologies eliminate water consumption for data centers at comparable cost

When should you avoid data center water consumption?

You're building data centers in water-stressed regions without alternative cooling. Your water reporting doesn't disclose per-facility consumption

What are alternatives?

Liquid cooling (direct-to-chip): Uses 90% less water than evaporative cooling — becoming standard for AI workloads. Nordic/cold climate locations: Free air cooling eliminates water use entirely — Iceland, Finland, Norway. Waste heat recovery: Use data center heat for district heating — turns waste into community benefit

Catalog

I011

Infrastructure

Data Center Water Consumption

HIGH(83%)

February 2026

4 sources

Context

Data centers generate enormous heat that must be dissipated. The most common cooling method uses water — evaporative cooling towers that consume millions of gallons per day. As AI drives explosive data center growth, water consumption is becoming a critical issue. Data centers are being built in regions already facing water stress. A single large data center can consume as much water as a small city, competing with agriculture, residential use, and ecosystems for an increasingly scarce resource.

Hypothesis

What people believe

“Cloud computing is environmentally efficient compared to on-premise infrastructure.”

Actual Chain

→

Water consumption at industrial scale(Large data center: 1-5 million gallons per day)

└

Evaporative cooling is the most cost-effective method — and the most water-intensive

└

A single Google data center uses as much water as 29,000 households

└

AI workloads generate more heat per rack — requiring more cooling

→

Data centers compete with communities for water in stressed regions(Data centers built in drought-prone areas for tax incentives)

└

The Dalles, Oregon: Google data center uses 25% of city's water

└

Mesa, Arizona: data centers approved despite severe drought conditions

└

Agricultural water rights challenged by data center demand

→

Corporate water reporting is opaque(Most tech companies don't disclose per-facility water use)

└

Aggregate water reporting hides impact on specific communities

└

'Water positive' pledges rely on offsets, not actual consumption reduction

→

Climate change worsens the problem(Hotter temperatures require more cooling, while water becomes scarcer)

└

Feedback loop: data centers contribute to warming, warming increases cooling needs

└

Drought conditions make evaporative cooling unsustainable in many regions

Impact

Metric	Before	After	Delta
Water per large data center	N/A	1-5 million gallons/day	Industrial scale
Water per AI query vs search	~0.3 mL (search)	~3-15 mL (AI query)	+10-50x
Tech company water consumption growth	Stable	+20-30% annually	Accelerating
Data centers in water-stressed regions	Few	Growing rapidly	Increasing conflict

Navigation

Don't If

•You're building data centers in water-stressed regions without alternative cooling
•Your water reporting doesn't disclose per-facility consumption

If You Must

1.Use air cooling or liquid cooling instead of evaporative cooling where climate permits
2.Build in regions with abundant water and cool climates (Nordics, Pacific Northwest)
3.Disclose per-facility water consumption transparently
4.Invest in water recycling and treatment for cooling water

Alternatives

Liquid cooling (direct-to-chip) — Uses 90% less water than evaporative cooling — becoming standard for AI workloads
Nordic/cold climate locations — Free air cooling eliminates water use entirely — Iceland, Finland, Norway
Waste heat recovery — Use data center heat for district heating — turns waste into community benefit

Falsifiability

This analysis is wrong if:

Data center water consumption decreases despite growing compute demand through efficiency improvements
Data centers in water-stressed regions do not measurably impact local water availability
Alternative cooling technologies eliminate water consumption for data centers at comparable cost

Sources

1.
University of California: Making AI Less Thirsty
Research quantifying water consumption of AI training and inference at data center scale
2.
AP News: Data Centers and Water
Investigation into data center water consumption in drought-affected communities
3.
Google Environmental Report
Google's water consumption increased 17% in one year, driven by AI data center expansion
4.
Microsoft Environmental Sustainability Report
Microsoft's water consumption increased 34% year-over-year, attributed to AI infrastructure growth

I016 A016 I001