AI’s future runs on water

The rise of artificial intelligence is transforming every industry, but it also creates enormous demand for digital infrastructure and natural resources. Data centers, the engines of this transformation, consume vast amounts of water and energy. 

A single hyperscale data center consumes up to 5 million gallons of potable water every day. In Phoenix, 58 centers together demand more than 170 million gallons daily, enough to ​​serve up to several hundred thousand households. 

This is the internet’s hidden water footprint, amplified by AI, cloud computing, and data-heavy services. Training a single large AI model in a Microsoft data center can require about 185,000 gallons of clean water. By 2027, AI-related data centers could consume 1.7 trillion gallons annually, nearly matching the domestic water use of some developed nations. 

Most data centers still rely on evaporative cooling, which consumes massive volumes and discharges chemical-laden wastewater. The challenge is not only scale but also geography. More than 40% of U.S. data centers are located in water-stressed basins. 

AI’s rapid growth demands a new approach. Water cannot become the bottleneck to the next chapter of human progress. 

Public concerns are already reshaping the industry: 

• Oregon: Google faced lawsuits over water secrecy. 

• Indiana: Amazon’s Project Rainier is under state scrutiny for allegedly draining wells while pumping millions of gallons per hour. 

• Georgia: Families near Meta’s complex report unusable wells. 

• Virginia: Utilities now require new data centers to secure their own water sources or adopt closed-loop systems. 

Investors are paying attention. Water use per AI training cycle is emerging as a core accountability metric, alongside carbon intensity. Communities are responding with moratorium requests. 

THE CALL FOR INDUSTRY LEADERSHIP 

The industry can no longer rely on incomplete data, inconsistent reporting, or distant offsetting schemes. Declaring a “water positive” target by some far-off date is no longer enough. Communities demand tangible action where the water is drawn. 

The technology exists today. Around the world, data center and cloud providers are proving that sustainability and scalability can coexist, with each breakthrough setting a new benchmark for what is possible. 

• Microsoft has deployed closed-loop systems in Arizona and Wisconsin, saving up to 125 million liters per site. 

• Amazon is building new centers with closed-loop treatment, recycling every drop used for cooling. 

• NVIDIA is partnering with Singtel to deploy next-generation liquid and immersion cooling systems designed to achieve industry-leading water efficiency in Singapore’s new AI data centers. 

The opportunity is clear: Water must be engineered into AI’s growth, not treated as an afterthought. 

THE TECHNOLOGIES DRIVING SUSTAINABLE AI 

Building a sustainable digital future requires bold adoption of both proven and emerging solutions that reduce environmental impact while enabling continued growth. The tools already exist. What we need now is the conviction to scale them. 

Smarter cooling technologies 

• Closed-loop and liquid cooling: Advanced systems can reduce water consumption by as much as 30 to 50% while maintaining the high-performance environment that AI workloads demand. 

• Water recycling at scale: Leaders like AWS plan to deploy treated wastewater at more than 120 data centers by 2030, setting a new baseline for responsible water use. 

AI-Driven optimization 

• Smart workload scheduling: By applying AI to manage computing loads, operators have shown they can ​​​​​​cut water consumption by a third without increasing carbon emissions. This type of efficiency breakthrough makes sustainability scalable. 

Alternative water sources 

• Seawater desalination: In coastal or arid regions, seawater offers an abundant alternative. Advanced desalination technologies convert it into a reliable cooling supply without burdening municipal drinking water systems. 

• High-value water reuse: Modern treatment technologies can transform sewage, brackish groundwater, and industrial effluent into high-quality process water, eliminating dependence on limited freshwater supplies. 

This is our approach at Gradiant, where our feedwater-agnostic treatment systems enable data centers to operate using seawater, wastewater, or other unconventional sources, reducing dependence on fresh supplies. By recycling blowdown and cooling tower reject, we achieve zero-liquid discharge and drastically reduce freshwater withdrawals, even in the largest hyperscale AI facilities. 

With the right technologies, sustainable AI data center growth can align with both environmental and business imperatives. 

AI’S GROWTH HINGES ON WATER 

The next era of AI will be defined by those who treat water as critical infrastructure. Companies that lead will gain faster permitting, avoid regulatory shocks and operational disruptions, and build lasting trust with the communities that host them. 

Water is not compliance. It is resilience. It is innovation. It is license to operate. 

AI’s future depends on leadership that recognizes water as the defining resource of our digital age, one that must be safeguarded through innovation rather than depletion. Advanced recycling, seawater desalination, and next-generation water treatment will be the pillars of responsible growth. The companies that act now will determine not only how AI grows but whether it grows responsibly, securing both digital progress and planetary resilience. 

Prakash Govindan and Anurag Bajpayee are the cofounders of Gradiant.