Thirsty Machines: How AI’s Water Consumption is Reshaping Data Centres
- Chris Howell
- Feb 7
- 4 min read
Updated: Jul 5
Introduction: The AI Boom and Water Demand
Artificial Intelligence (AI) is expanding at an unprecedented rate, with data centres emerging across the UK to support growing computational demands. From chatbots to large language models, AI requires significant processing power—power that generates heat and demands efficient cooling solutions.
The challenge? Data centres consume massive amounts of water for cooling. As the UK pushes forward with its ambitious AI growth strategy, a crucial question arises: Can AI help address its own water challenge?
The Reality of the UK’s Water Scarcity
At first glance, the UK shouldn’t face water shortages. After all, it’s an island surrounded by water and has a reputation for rainy weather. However, the issue isn’t a lack of water but access to fresh, clean water in the regions where it is most needed.
Freshwater availability is limited – Data centres require clean, non-corrosive water for effective cooling.
Water distribution is uneven – The southeast of England, one of the driest regions, hosts a significant number of data centres.
Climate change is reducing supply – Hotter summers and shifting rainfall patterns are depleting groundwater sources and increasing the risk of droughts.
Competing demands – Agriculture, manufacturing, and urban populations already place significant stress on water resources.
Even today, London and the southeast regularly experience water stress, a problem likely to intensify as AI infrastructure expands.
Why Seawater Isn’t a Simple Solution
If the UK is surrounded by seawater, why not use it for cooling? While possible, several barriers exist:
Seawater is highly corrosive – Standard cooling systems aren’t designed to handle saltwater.
Desalination is costly – Removing salt from seawater is energy-intensive and not yet widely adopted.
Environmental impact – The desalination process generates brine waste, which can harm marine ecosystems.
Some coastal data centres could explore seawater cooling, but inland facilities remain dependent on freshwater resources.
Fancy a Drink? How AI’s Water Consumption Compares to Humans
Water is essential for life, but how does AI’s water consumption compare to that of humans?
An average person drinks about 2-3 litres of water per day – that’s around 1,000 litres per year.
Traditional data centres (without AI) can use millions of litres annually, depending on their size and cooling systems.
AI-powered data centres consume even more, with Google alone using 15 billion litres of water in its data centres in 2022.
A single AI model training cycle can use millions of litres—far exceeding the lifetime drinking water needs of thousands of people.
While humans need water for survival, AI’s thirst is growing rapidly as more businesses integrate AI-driven solutions. Balancing AI expansion with sustainable water use is crucial to prevent unnecessary strain on global freshwater supplies.
AI’s Water Footprint and the Path to Sustainability
AI’s exponential growth brings with it a significant water footprint. Every AI-powered interaction, training session, and computational process contributes to cooling demands.
Google reported using 15 billion litres of water in its data centres in 2022 alone.
Microsoft and other tech giants have been criticized for excessive water consumption.
A single AI model training cycle can consume millions of litres of water.
However, AI isn’t just the cause of the problem—it can also be a key part of the solution.
Solutions: AI Leading the Way in Water Efficiency
The UK government acknowledges the need for sustainable AI infrastructure, particularly in addressing environmental challenges such as water consumption in data centres. Future policies may incentivize innovations in water-efficient cooling methods and sustainable AI deployment. While AI is seen as a key driver of economic transformation, water conservation has yet to be fully integrated into AI-related infrastructure planning.
Rather than exacerbating water shortages, AI can be leveraged to optimize water management and improve sustainability. Potential solutions include:
AI-Optimized Cooling Systems – AI-driven analytics can predict cooling needs and reduce excess water use through real-time adjustments.
Use of Recycled Wastewater – Data centres can repurpose treated wastewater instead of relying on potable fresh water.
Investment in Seawater Cooling – Coastal data centres could explore saltwater-resistant cooling infrastructure.
Improved AI Model Efficiency – More efficient AI training reduces energy and water demands.
Predictive AI for Water Management – AI-driven forecasting can help mitigate drought risks and optimize water conservation efforts.
Government Policies & Industry Standards – The UK needs clear regulations and incentives to encourage sustainable AI infrastructure.
Conclusion: AI Can Help Solve Its Own Water Challenge
AI is undoubtedly shaping the future, but its growth must be sustainable. The good news? AI can play a role in solving the water crisis by driving innovation in cooling technology, predictive analytics, and efficiency improvements.
By adopting sustainable infrastructure, responsible AI development, and forward-thinking policies, the UK can ensure that its AI ambitions align with environmental responsibility.
(This article was inspired by a report from the BBC on the potential water challenges faced by UK data centres. You can read their original coverage here.)
