Why Investors Are Pressing Tech Giants on Water Use
Investors are demanding answers on water use in data centers. Explore the implications for sustainability and investment decisions.
Water has become one of the most contentious resources in the technology industry — and investors are paying serious attention.
When shareholders at Amazon, Microsoft, and Google began formally pressing these companies for detailed disclosures on data center water consumption, it wasn't a fringe environmental protest. These were institutional investors — the kind managing pension funds and endowments — asking pointed, specific questions about operational risk. That distinction matters. This isn't advocacy; it's risk analysis.
The underlying concern is straightforward: data centers use an enormous amount of water to cool their servers, and that demand is growing faster than most people realize. But the full story — why it's happening, what the actual numbers look like, and who ultimately bears the cost — is worth understanding carefully.
The Scale of the Problem Is Larger Than the Headlines Suggest
A single large hyperscale data center can consume anywhere from 1 million to 5 million gallons of water per day. To put that in context, that's roughly equivalent to the daily water use of a city of 10,000 to 50,000 people — drawn from the same municipal or watershed supply those communities depend on.
Data center water use isn't a niche environmental issue anymore; it's a resource competition that communities and utilities are increasingly losing.
Microsoft disclosed that its global water consumption grew by 34% between 2021 and 2022, reaching approximately 6.4 million cubic meters. Google's water use exceeded 5.6 billion gallons in 2022. Amazon has been notably less forthcoming with granular figures — which is precisely why investors are asking the questions they're asking.
The mechanism driving consumption is basic thermodynamics. Servers generate heat. That heat has to go somewhere. The dominant cooling method — evaporative cooling — removes heat by evaporating water, and that water doesn't come back. It's consumptive use, not recycled use. Every AI training run, every cloud storage request, and every streamed video generates heat that, in most facilities, gets managed by evaporating water drawn from local sources.
Why Data Centers Are in a Different Category Than Other Industries
Industrial water users — agriculture, manufacturing, power generation — have faced regulatory scrutiny over water consumption for decades. Data centers largely escaped that scrutiny because, individually, they looked small. A warehouse-scale building with a modest visible footprint didn't trigger the same alarm as a coal plant or a semiconductor fab.
That calculus is changing. The aggregate footprint of hyperscale cloud infrastructure has reached a scale where it registers as a major water consumer at the regional level. In water-stressed areas like the American Southwest, Northern Virginia's suburban sprawl, or parts of Western Europe, data center clusters are now competing directly with agriculture, municipal water systems, and environmental flow requirements.
The difference between a data center and a steel mill, from a water management perspective, is that nobody questions whether we still need steel mills.
The demand case for cloud computing is not in question — it's the fastest-growing segment of enterprise infrastructure spending. But that growth trajectory makes the water math increasingly uncomfortable. AI workloads, in particular, are significantly more compute-intensive than standard cloud operations, which translates directly into higher cooling loads and higher water demand per unit of useful output.
What Investors Are Actually Demanding
The shareholder resolutions filed against Amazon, Microsoft, and Google weren't vague appeals to be more sustainable. They asked for specific disclosures: facility-level water consumption data, water risk assessments tied to geographic location, and clear strategies for reducing consumptive water use over defined timelines.
That level of specificity signals a maturing of ESG investment pressure. Early-era ESG activism often asked companies to "commit to sustainability." Current institutional investor pressure is asking for auditable data that can be used to model operational and reputational risk.
The responses from the three companies have varied meaningfully. Microsoft has been the most forthcoming, publishing detailed sustainability reports and committing to being "water positive" by 2030 — meaning it aims to replenish more water than it consumes. Google has made similar commitments and invested in watershed restoration projects. Amazon has lagged on transparency, offering aggregate figures without the facility-level granularity investors have requested.
That gap in disclosure is itself a red flag for sophisticated investors. When a company resists disclosing data that its peers are voluntarily publishing, the question isn't whether there's something to hide — it's what the risk profile looks like when the data eventually surfaces.
The Financial Risk Is Real, Not Theoretical
Water risk for data centers breaks down into three distinct categories, and none of them are speculative.
First, there's regulatory risk. Several U.S. states and European jurisdictions are actively tightening water allocation rules. A data center that secured water rights in 2015 under one regulatory regime may find those rights restricted or subject to new fees within a decade. Arizona, already contending with a shrinking Colorado River allocation, has begun scrutinizing large water users in ways that will affect data center permitting.
Second, there's operational risk. Water stress events — droughts, supply disruptions — can force facilities to shift to less efficient cooling methods or curtail operations. A facility designed around consistent access to low-cost evaporative cooling has a fundamentally different risk profile in a drought year.
Third, and perhaps most directly relevant to investors, there's stranded asset risk. A data center sited in a location that becomes water-constrained within its operational lifetime — typically 20 to 30 years — faces significant capital loss. The infrastructure can't easily move.
What Sustainable Data Centers Actually Look Like
The good news is that technical solutions exist. The harder news is that they involve real tradeoffs in cost and efficiency.
Air-cooled systems eliminate water consumption but require more energy and work less efficiently in high ambient temperatures. Closed-loop liquid cooling — where coolant circulates through server racks and rejects heat without evaporation — can dramatically reduce water use but requires significant infrastructure investment and isn't yet standard across hyperscale deployments. Microsoft has experimented with underwater data centers and geothermal cooling; neither is ready for mainstream deployment at scale.
The most practical near-term strategy is something the industry calls Power Usage Effectiveness (PUE) optimization paired with Water Usage Effectiveness (WUE) targets. WUE measures liters of water consumed per kilowatt-hour of IT load. The industry average hovers around 1.8 L/kWh; best-in-class facilities achieve below 0.5 L/kWh. That gap represents an enormous range of outcomes — and the facilities being built today will operate for decades.
Siting decisions made in 2024 will determine water risk exposure through 2050, which is exactly the time horizon that long-term investors are modeling.
Location strategy is increasingly driving the conversation. There's genuine interest in siting data centers in cooler climates — the Nordic countries have been a significant beneficiary — where ambient air temperatures reduce cooling loads and water consumption simultaneously. Iceland, with abundant geothermal energy and cold air, has become a serious destination for certain workloads. The tradeoff is latency and connectivity; not every workload can tolerate distance from end users.
Where This Goes Next
Investor pressure on data center water use is not going to decrease. The combination of AI-driven demand growth, tightening water regulation in key markets, and increasing institutional focus on physical climate risk creates durable structural pressure on the industry to improve disclosure and reduce consumption.
The companies that get ahead of this — publishing auditable, facility-level water data, investing in low-water cooling technology, and building siting strategies that account for 30-year water availability — will find themselves better positioned with institutional capital and better protected against regulatory disruption.
The companies that continue to resist granular disclosure should expect shareholder resolutions to escalate. Institutional investors have demonstrated in other sectors — oil and gas, agriculture, apparel — that when disclosure requests are ignored, they move toward binding votes, board-level accountability measures, and eventually divestment screening.
Water is infrastructure. In a world where AI training runs consume the computational equivalent of small cities, the resource inputs that make computation possible deserve the same scrutiny as the technology itself. The investors pressing Amazon, Microsoft, and Google on this aren't being idealistic. They're doing their jobs.
Explore more about how InfraSale Marketplace is addressing these challenges.