In recent months, alarming headlines have surfaced claiming that AI bots consume millions of liters of water every day. But wait — how does software drink water?
The answer is more nuanced than it sounds. While AI systems like ChatGPT, Gemini, and other large language models (LLMs) don’t directly consume water, the infrastructure that powers them — namely data centers and energy production systems — certainly does.
Let’s unpack the hidden water footprint of artificial intelligence.
The Hidden Infrastructure Behind AI
Every time you ask a chatbot a question, generate an image with AI, or train a machine learning model, you’re tapping into vast networks of data centers. These buildings are filled with power-hungry servers, all of which generate massive amounts of heat. To keep these systems operational and prevent overheating, they must be cooled — and that’s where water comes in.
Two Main Sources of Water Use in AI:
- Data Center Cooling
- Most data centers use water-based cooling systems to keep servers at safe operating temperatures.
- This often involves evaporative cooling, where water is used to absorb heat and then evaporated, or closed-loop systems that circulate chilled water.
- In hot or dry climates, even more water is needed to maintain efficient cooling.
- Electricity Production
- AI systems consume a huge amount of electricity.
- Many power plants — especially thermoelectric and hydroelectric plants — consume or evaporate large quantities of water during the electricity generation process.
- So even indirectly, AI is “drinking” water every time it runs.
How Much Water Are We Talking About?
Let’s take Google as an example. In 2022, it was revealed that:
- Some of Google’s data centers use up to 2.1 million liters (550,000 gallons) of water per day.
- On an annual scale, the company used over 20 billion liters of water across its facilities.
As AI models get larger and more powerful, their demand for energy and cooling continues to grow — making their water footprint even more significant.
Is This Harmful to the Planet?
Yes, the environmental implications of such high water use can be very serious, especially in water-stressed regions.
Key Environmental Concerns:
1. Water Scarcity
Many data centers are located in or near drought-prone areas, where water is already a limited resource. For example:
- In 2023, Google faced backlash for planning a new data center in Uruguay, which was experiencing its worst drought in decades.
- Local residents and activists protested, fearing the facility would worsen water shortages for agriculture and households.
2. Depletion of Freshwater Sources
Most data centers rely on fresh, potable water. Excessive withdrawal can lower groundwater levels, dry up rivers and lakes, and disrupt local ecosystems.
- Especially in evaporative cooling systems, water is lost permanently to the atmosphere.
3. Carbon Footprint and Energy Use
Cooling systems that rely on electricity from fossil fuels not only use water but also contribute to carbon emissions. So, high water use often goes hand-in-hand with a larger climate impact.
4. Ecological Disruption
Drawing large amounts of water from natural sources can affect fish populations, wetland ecosystems, and biodiversity.
5. Wastewater and Thermal Pollution
Water discharged from data center cooling systems is often hot or contains chemical residues. If not treated properly, this pollutes rivers and lakes, harming aquatic life.
So, What Are Companies Doing About It?
Big tech companies are aware of these issues and are actively working on sustainability strategies.
Some of the solutions include:
✅ Using Renewable Energy
Shifting data centers to solar, wind, or hydroelectric power reduces the indirect water use and carbon footprint associated with traditional power sources.
✅ Water Recycling and Air Cooling
- Some companies are experimenting with air-based cooling systems, which drastically reduce water use.
- Others invest in recycling greywater or treating and reusing cooling wastewater.
✅ Smarter Location Choices
Placing data centers in regions with plentiful water resources and cooler climates can reduce both water and energy demand.
✅ AI-Optimized Cooling
Ironically, AI itself can be used to optimize AI. For example:
- Google’s DeepMind helped reduce energy used for cooling its data centers by up to 40% by adjusting cooling settings in real time based on environmental data.
✅ Transparency and Reporting
Publicly disclosing water usage and sustainability goals helps companies stay accountable and allows public oversight.
Fun Fact: Water + AI = Science, Too!
Not all water-related AI stories are negative. For example, in 1992, nearly 29,000 rubber ducks were accidentally released into the Pacific Ocean from a cargo ship. Scientists later used their drifting patterns to study ocean currents — aided by AI models decades later to simulate those same movements.
It’s a small but fascinating reminder that AI and water are deeply interconnected, in both environmental and scientific contexts.
Final Thoughts: Balancing Progress with Responsibility
As artificial intelligence continues to grow in scale and impact, so does its environmental footprint. While the benefits of AI are undeniable — from healthcare to education to automation — its resource consumption must be acknowledged and addressed.
🌍 The challenge isn’t to stop AI, but to make it smarter — and greener.
That means:
- Building more efficient infrastructure
- Using AI to monitor and reduce its own footprint
- Creating global standards for water and energy use in tech
