AI Feels Invisible. Its Water Use Isn't.
Typing a question into ChatGPT takes only a few seconds.
There's no smoke, no factory, and no obvious sign that anything physical is happening behind the scenes. It feels like one of the cleanest technologies ever created.
But every response is powered by massive data centers filled with thousands of high-performance processors working around the clock. Those processors generate enormous amounts of heat, and keeping them cool often requires one resource that's becoming increasingly precious: water.
As artificial intelligence spreads into search engines, office software, smartphones, and businesses, experts are beginning to ask a new question:
Could AI become one of the next major consumers of the world's water resources?
Read More
Why Does AI Need Water?
The AI model itself doesn't «drink» water.Instead, the water is used to cool the servers running the model.
Modern AI systems rely on powerful graphics processors (GPUs) that perform billions of calculations every second. During training and even while serving millions of user requests, these chips generate significant heat.
Without cooling systems, temperatures would rise quickly enough to damage expensive hardware or reduce performance.
Many data centers remove this heat using cooling technologies that depend on water, particularly evaporative cooling systems.
Did Researchers Really Estimate Half a Liter of Water for 20–50 Prompts?
This claim has circulated widely across social media—and it does have a scientific origin.
In 2023, researchers from the University of California, Riverside, and the University of Texas at Arlington estimated that 20 to 50 ChatGPT prompts could indirectly consume about 500 milliliters (17 ounces) of water, depending on where and how the AI service is operating.
However, there's an important detail that often gets left out.
The figure is not a fixed measurement.
It is a research estimate based on several variables, including:
- The location of the data center
- Local climate and temperature
- Cooling technology
- Electricity generation
- Server efficiency
In other words, not every ChatGPT conversation consumes the same amount of water.
AI Infrastructure Is Already Becoming More Efficient
The AI industry has changed rapidly since those early estimates.
Major technology companies are investing billions of dollars to reduce the environmental impact of their data centers.
Microsoft, for example, has announced new cooling designs that significantly reduce water intensity in next-generation facilities. Other companies are increasingly adopting closed-loop cooling systems, recycled water, and liquid cooling technologies to improve efficiency.
Recent analyses also suggest that the water footprint of a typical AI request may be considerably lower than earlier estimates in some modern data centers.
The technology is evolving almost as quickly as AI itself.
Why Microsoft's Water Consumption Jumped 34%
One statistic that surprised many observers came from Microsoft's own sustainability reporting.
The company reported a 34% increase in water consumption during a period of rapid cloud infrastructure expansion.
The primary reason wasn't office buildings or manufacturing—it was the rapid growth of cloud computing and AI infrastructure needed to support services used by millions of people every day.
As AI demand continues to rise, infrastructure must expand with it.
Could AI Use Trillions of Liters of Water?
Some researchers believe global AI-related data centers could consume trillions of liters of water annually over the coming years if current growth trends continue.
That doesn't mean the outcome is inevitable.
Future water demand will depend on several factors:
- Advances in cooling technology
- More energy-efficient AI chips
- Smarter data center design
- Increased use of recycled water
- The overall growth of AI services worldwide
The industry's environmental footprint is still being written.
AI Isn't the Only Digital Technology Using Water
It's easy to single out artificial intelligence, but AI isn't alone.
Nearly every major online service depends on data centers, including:
- Search engines
- Cloud storage
- Video streaming
- Social media platforms
- Online gaming
- Business software
What makes AI different is the enormous amount of computing power required to generate intelligent responses in real time.
That extra computing power also increases demand for cooling.
How Tech Companies Are Trying to Reduce Water Use
The world's largest technology companies know that water has become a strategic issue.
To reduce their environmental impact, many are investing in:
- Direct liquid cooling
- Closed-loop water systems
- Recycled water instead of drinking water
- More energy-efficient AI processors
- Data centers built in cooler climates
The goal is simple: deliver more computing power while consuming fewer natural resources.
Should Everyday Users Be Worried?
For individual users, asking a few questions to ChatGPT won't have a noticeable environmental impact.
The bigger issue is scale.
Every day, billions of AI requests are processed around the world. When those requests are combined, the demand for electricity and cooling becomes substantial.
That doesn't mean AI is inherently bad for the environment.
Instead, it highlights a growing challenge: making the next generation of artificial intelligence both smarter and more sustainable.
