As they face increasing backlash over their resource consumption, major data center operators are scrambling to prove they’re not a drain on the environment, or, at least, not as much of one as their competition.
Amazon has published some bold new claims to this end: The tech giant says it has achieved a 52% improvement in water efficiency over the last 5 years, and says its data centers are 7x more water-efficient than the industry average.
This, the company says, is thanks to a mix of innovative methods, including free air and evaporative cooling, and increased temperature thresholds.
The announcement underscores the importance of disclosure in the AI era, and signals that technology is no longer the sole differentiator, said Sanchit Vir Gogia of Greyhound Research. “Water efficiency has become a front in hyperscale competition, not a footnote.”
How Amazon is reducing its water consumption
Amazon’s global data center operations used 0.12 liters of water per kilowatt-hour (L/kWh) in 2025, compared to the industry average of 0.84 L/kWh, a 7x lead, the company says.
Its closest competition was Microsoft, which used 0.27 L/kWh in 2025, dropping from .30 in 2024. Google seems to be the heaviest user in recent years (averaging 1.15 L/kWh), while Meta has hovered around .20 L/kWh.
Amazon employs a multiple methods to garner these results, it says.
Firstly, roughly 90% of the time the company uses “free air cooling,” pulling outside air into data centers, letting the air absorb heat, then pumping it back outside, with no water required. Amazon likens it to opening the windows on a summer evening rather than running the air conditioner.
When temperatures rise, Amazon uses evaporative cooling. Water is sprayed onto an absorbent medium and air is allowed to flow through it. The water evaporates and pulls heat from the air, cooling it by 5 to 10 degrees Fahrenheit.
Further, the company has been deliberately raising the thresholds at which its centers operate, designing servers that can tolerate more heat, thus reducing water needs. After a few years of “iteration, learning, and adjusting,” Amazon has raised the operating temperature to 85° F.
Reclaiming water, partnering on community projects
In 2025, Amazon says it returned 3 US gallons of water to local communities for every 4 gallons used, and is 75% of the way towards its goal of being water positive by 2030, where every gallon used will be returned in kind.
The company uses reclaimed water sourced from wastewater treatment plants, as opposed to potable water, across 130 of its data centers, with 26 facilities using this method exclusively. Additionally, it is helping communities develop reclaimed water programs that could return more than 5.8 billion US gallons annually.
Amazon is particularly focusing on areas where water is scarce, and aims to partner with communities “to ensure our water stewardship creates local benefits that they want to see,” the company says.
Not hollow claims, but there are nuances
“The gains are certainly real and the engineering is legitimate, so Amazon earns credit here,” said Matt Kimball, VP and principal analyst at Moor Insights & Strategy.
The 0.12 figure is Water Usage Effectiveness (WUE), which measures the water used at the data center per kilowatt-hour of IT load. The metric was introduced by Green Grid and is widely accepted as the standard, he explained.
Still, it can be difficult to be precise in these measurements, and there are nuances: Is a data center operator only reporting WUE? Do they count the water used to generate electricity (most actually don’t)? Is reclaimed water being counted in the waysame as potable?
But “to Amazon’s credit, they aren’t just shifting the burden elsewhere,” Kimball noted.
Free air cooling and higher operating temperature thresholds reduce both water and energy consumption, and Amazon’s power usage effectiveness (PUE), the standard metric for measuring data center efficiency, sits around 1.15.
“This means they’re not achieving this strong water number at the expense of a weak energy number,” Kimball said.
Not a secret recipe
While the gains are impressive, the techniques Amazon highlights are increasingly becoming standard practices, Kimball pointed out.
“Amazon is on the leading edge, but it’s not a secret recipe,” he said. What sets the company apart is scale, execution, facility design, geographic mix, and its aggressive pursuit of energy goals.
Others are doing the similar things, if through different avenues: Microsoft is investing in closed-loop cooling systems that dramatically reduce evaporative water loss. Google is heavily focused on reclaimed water and using AI to optimize data centers. Meta has long relied on outside-air cooling. And overall, the industry is moving toward liquid cooling for dense AI deployments, “which changes the water equation again,” said Kimball.
One of the big variables is location: Climate influences water efficiency, so where a company builds its infrastructure is as important as its cooling methods. Further, power-consumptive AI changes the discussion, he emphasized; traditional enterprise workloads and dense AI training clusters create very different thermal profiles.
“The industry is still working through what the long-term water and energy balance looks like as AI infrastructure scales,” Kimball said.
The disclosure arms race
As Gogia noted, what separates operators now is disclosure.
For instance, Microsoft has committed to publishing water data for every US data center region, while Equinix reports a portfolio figure of 0.91 L/kWh and 1.41 L/kWh for its evaporative-cooled sites, which is a “boundary discipline most of the market has yet to adopt,” said Gogia.
The next phase of cloud competition will be won on transparency, not just thermodynamics, he contended. “Operators that disclose region by region will expand faster, permit easier and litigate less.” In fact, his firm expects water disclosure to become a standard line item in cloud and AI infrastructure RFPs within 12 to 24 months.
CIOs should demand region-level withdrawal and consumption data, source-water mix, cooling architecture by region, drought contingency plans, and a “clean separation between operational efficiency and replenishment accounting,” Gogia advised.
Meanwhile, the timing of this reporting tells its own story: Amazon published these figures two days after its home city Seattle put a one-year freeze on new large data centers, citing water use as a big concern. Further, more than 70 US jurisdictions now have temporary or permanent restrictions, and the European Commission is preparing efficiency standards with water criteria.
So, this can be construed as a sustainability report as much as a “social-license counteroffensive,” said Gogia.
AI is changing the game entirely
Buyers should care about these developments for practical reasons beyond environmental, social, and governance (ESG) reporting, Kimball said.
First, water is becoming a real constraint on where data centers can be built, and communities are pushing back on development. This can impact capacity, expansion timelines, and operating costs in high-demand markets, he said.
Second, a cloud provider’s water footprint increasingly becomes part of a customer’s own sustainability profile. Therefore, organizations with environmental reporting requirements or water-related goals are paying closer attention to the resource consumption of the infrastructure supporting their apps and AI workloads.
Ultimately, Gogia noted, data centers are no longer invisible infrastructure: They are becoming contested civic infrastructure.
“The winners will treat water as a shared public constraint rather than a line item in sustainability theater,” he said. “The future of AI infrastructure will be decided as much by resource stewardship as by engineering capability.”