When the dry, hot months arrive, households turn on their taps, putting local water infrastructure under peak pressure. Increasingly, so do their new neighbors: data centers powering a worldwide surge in AI.
These data centers need the most water at exactly the moment when communities can spare the least. For operators, that means managing high-demand cooling periods, meeting uptime targets and navigating stricter local regulations. Around 40% of data centers are already in water-stressed regions. And by 2050, their demand for water will rise by an estimated 272%, according to the Watering the New Economy report published by Global Water Intelligence (GWI) and Xylem.
But there is a way data centers can avoid a trade-off with communities over a shared, essential resource. A way to advance AI innovation and protect people. This path requires a fundamental shift toward reusing more water, plugging leaks and forging partnerships that help municipalities modernize their infrastructure. The report refers to this as a broad-based, global water transition – and data centers are at the heart of it.
AI growth starts with water
The Watering the New Economy report gives us a comprehensive assessment of how AI is reshaping global water use, along with a fact-based framework to rethink how we manage and protect the water that AI innovation depends on. We now have a blueprint for transforming water from AI’s hidden vulnerability into a source of resilience.
Water underpins the entire AI supercycle, from data centers and semiconductor manufacturing to power generation. Cooling drives almost all the onsite water demand at data centers.
Sector-wide cooling efficiency is improving and will continue to improve by as much as 46% by 2050, according to the report. Keeping servers from overheating is also a far less water-intensive process than fabricating computer chips (which requires filtered, ultrapure water) or generating offsite electricity.
But the reality remains. Aging water infrastructure is struggling to keep up with demand from both communities and a rapidly expanding network of AI data hubs – especially during summer months. Data centers are increasingly being built in dry or rural areas, where land and power are cheap, but water is far from abundant. And unlike fabrication plants, data centers consume (through evaporation) most of the water they draw, with even the most efficient cooling systems discharging just 15% of their withdrawals.
From strain to solutions
These converging risks help explain why giant tech companies – from Amazon to Microsoft, from Alphabet to Meta – are committing to ambitious ‘water positive’ goals by 2030, undertaking to replenish more water than they consume. There are growing industry calls for tech companies to be ‘good neighbors’ and to pursue AI expansion in a way that meets community needs.
Pushback from residents can lead companies to relocate, as was the case in Arizona where already-scarce water supplies were threatened by a semiconductor packaging plant. In Chile, protests and political debates have flared up over AI infrastructure expansion.
But states like Arizona also offer a model for the solution. In Chandler, Intel funded a water treatment facility that’s achieving 96% water recovery and reuse. Treated water like this can be sent back to communities, agriculture, and neighboring industries.
The treatment plant is owned and operated by a local utility. This partnership allows the City of Chandler municipality to improve public infrastructure, while strengthening the region’s water supply chain, which Intel relies on.
A moment to act
Strengthening shared systems also means stopping leaks. According to our report, global wastewater volumes total around 320 trillion liters a year, and up to 100 trillion liters lost to leaking pipes could be recovered – far exceeding the additional water demand the AI supercycle is expected to add in the next quarter century.
In Mexico for example, Amazon has partnered with local utilities to deploy smart water management solutions that use data, analytics, advanced pressure management and real-time leak detection.
Collaboration and targeted investment in intelligent infrastructure deliver resilience and protect community water systems. Operators are exploring innovative air-cooling and water reuse solutions at data centers. Some are even storing rainwater to ease pressure on local water systems.
These efficiency gains help decouple AI expansion from water availability. Another crucial element of the water transition is optimizing the energy mix to support the expected growth of renewable energy. Offsite power generation for data centers and fabrication plants accounts for roughly half (54%) of AI’s current footprint. The GWI and Xylem report estimates renewables could save just over 100 trillion liters of water by 2050.
A clear imperative for water leadership
As AI drives the fastest industrial expansion in history, the spotlight has been on its hunger for energy – not its thirst for water. But with AI-driven water demand set to surge nearly 130% over the next 25 years, water security is increasingly becoming a strategic priority.
Leaders across the sector now have the chance to build systems that protect communities and secure long-term capacity. We can choose a path that supports AI innovation and protects water for communities – one built on reuse, efficiency and shared investment. Ultimately, the new AI economy will require a successful water transition. Handled responsibly, this water transition can become a foundation for data center resilience – supporting uptime, permitting, community trust and long-term performance. The tools exist. What’s needed now is joint commitment and cross-sector collaboration.
