Microsoft Data Center Water Use to Double by 2030

Microsoft now projects its data center water use will more than double by 2030 despite water-use intensity improvements, as expansion accelerates in regions already facing shortages.

The projection contradicts its pledge to operate all data centers without wasting water by 2026. The company’s water use baseline is from 2020 levels, with the expected doubling occurring alongside rapid AI datacenter expansion.

Microsoft’s Response

Despite these projections, Microsoft announced its Community-First AI Infrastructure initiative aimed at strengthening local water systems. The company committed to a 40 percent improvement in datacenter water-use intensity by 2030 and pledged to replenish more water than it withdraws from local districts.

Microsoft also launched a new AI datacenter design featuring a closed-loop cooling system that eliminates potable water use entirely. While the 40 percent intensity improvement positions Microsoft to reduce water consumption per computation, absolute usage still doubles. This reflects the scale at which AI infrastructure is expanding.

Operational efficiency gains fail to offset the sheer volume of new datacenter capacity required to meet enterprise AI demand. Framing these efforts as community investments, Brad Smith, Microsoft’s Vice Chair and President, emphasized the company’s commitment to strengthen rather than strain local water resources.

“Our commitment ensures that our presence will strengthen local water systems rather than burden them. We’ll do this by reducing the amount of water we use and by investing in local water systems and water replenishment projects.”

Brad Smith, Vice Chair & President at Microsoft (via Microsoft blog)

Building on this commitment, Microsoft pointed to existing partnerships. These include the Quincy Water Reuse Utility in Washington state that treats and recirculates datacenter cooling water. The company also highlighted $25 million in infrastructure investment for water and sewer improvements near its Leesburg, Virginia facility.

These projects demonstrate Microsoft’s efforts to address immediate community needs while scaling AI infrastructure.

Industry Scale

Yet Microsoft’s challenges reflect broader trends across the datacenter industry. Amazon, Microsoft, and Google are operating 38 active datacentres in water-scarce areas, with 24 more under development. Microsoft said 42% of its water came from areas with water stress in 2023.

Across the industry, US data centers consumed more than 17 billion gallons of water in 2023. The largest hyperscaler facilities use up to 5 million gallons on hot days.

The concentration of 62 datacenters in water-stressed regions indicates a fundamental mismatch between infrastructure placement and resource availability. For communities already managing limited water supplies, this positions tech companies as competing directly with residential and agricultural users during shortage periods.

This pressure intensifies as AI workloads accelerate growth. Shaolei Ren, a computer engineer at University of California, Riverside, described AI-driven datacenter environmental costs as “the fastest-growing sector” in terms of environmental impact. This rapid expansion reflects both the computational demands of training large AI models and the cooling requirements needed to maintain optimal operating temperatures for processing hardware.

Beyond water consumption, data centers already account for about 4 percent of electricity consumed in the United States, according to Pew Research. That figure is projected to rise to between 7 and 12 percent by 2028 according to industry estimates.

The International Energy Agency estimates datacenter electricity demand will more than triple by 2035, from 200 to 640 terawatt-hours per year. This tripling of electricity demand far exceeds projected water consumption increases, but the two resources remain coupled through cooling requirements.

As datacenter density increases to support AI workloads, cooling systems must dissipate proportionally more heat, creating compounding pressure on both power grids and water infrastructure in host regions.

Competitive Context

Across the industry, tech companies face similar pressures to balance AI growth with environmental impact. Amazon, Google, and Microsoft each vowed to replenish more water than they consume by 2030.

Amazon Web Services expanded recycled water use to 120 locations across the United States. The expansion is expected to preserve over 530 million gallons of fresh drinking water annually.

However, water replenishment pledges face inherent limitations. As Aaron Wemhoff, an energy efficiency specialist at Villanova University, explained: “Carbon is a global problem – water is more localised.”

Unlike carbon offsets, water consumed in one region cannot meaningfully benefit communities hundreds of miles away facing their own shortages.

Replenishing water in one watershed does little to address depletion in another, particularly when datacenters concentrate in areas already experiencing water stress. Compounding these challenges, corporate transparency remains limited, complicating efforts to assess true environmental impact.

Priya Donti, a machine learning researcher at Massachusetts Institute of Technology, identified data transparency as “a huge issue” with hyperscalers failing to provide granular usage information.

Without detailed disclosure, communities struggle to evaluate datacenter operations’ impact on local water supplies as AI infrastructure expands. This challenge intensifies the tension between Microsoft’s conservation pledge and its projected doubling of water consumption by 2030.