AI Water Usage in Data Centers: How Machines Are Cooled and How Much Water They Use
For Metis Consulting Services
Written by Michael Bronfman
September 15, 2025
This week in The Guard Rail, Metis Consulting Services' thought leadership blog, we're taking a look at a hidden environmental cost of our digital lives. While the pharmaceutical industry meticulously manages every drop of liquid in manufacturing processes, another sector, the data center industry, is gulping down millions of gallons of water a day to keep our modern world digitally running. We'll explore how these massive server farms are cooled and why their water consumption is becoming a significant concern, creating a new kind of "liquid asset" problem that requires a creative and sustainable solution.
Water Management Reality
Modern life relies on powerful computer systems that store information, process data, and maintain digital services. These large computer facilities are called data centers. Every time someone uses a search engine, streams a video, or stores a photo online, data centers are at work behind the scenes. While most people consider the electricity required to keep these machines running, fewer people think about another resource that data centers consume: water.
Water is used mainly for cooling. Computers generate heat when they operate, and if they become too hot, they may stop working or fail completely. Cooling systems keep machines at the right temperature. In many cases, water plays a central role in this process. As the demand for computing continues to grow rapidly, the amount of water used by data centers is becoming a significant environmental concern.
This essay explains how water is used to cool machines, why water is chosen, how much water is consumed, and what can be done to reduce water use.
Why Cooling Is Needed
Computers generate heat because electrical energy is transformed into thermal energy as circuits work. The more powerful the computer, the more heat the data center releases. Thousands of servers operate simultaneously in a single building. Without cooling, the heat would build up and damage the equipment.
The cooling process maintains a stable temperature, protects equipment, and enables data centers to operate continuously around the clock. Cooling also affects efficiency. A data center that runs too hot requires emergency shutdowns, which wastes electricity and can interrupt services.
How Data Centers Are Cooled
There are various methods to cool data centers, but many of them involve the use of water.
Air Cooling
Some data centers use outside air to reduce heat. They blow cool air through server racks, pushing hot air out. This system works better in cooler climates, but it is less efficient in warm regions.
Chilled Water Cooling
Many data centers use chilled water systems. Large chillers cool water, and then cold water circulates through pipes to absorb heat from the servers. The warmed water goes back to the chillers, where it is cooled again..
Cooling Towers
Cooling towers release heat from water by allowing it to evaporate. Water is sprayed into the air, and as some of it evaporates, the remaining water cools. This cooled water is reused again in the system.
Direct Liquid Cooling
Some advanced systems pump water or special liquids directly to the computer chips. This method reduces the need for massive air systems and can be more efficient, but it still requires a supply of water.
Why Water Is Used
Water is an effective cooling material because it has a high heat capacity. This means it can absorb and carry away large amounts of heat. Water is also widely available and cheaper than many alternatives.
However, water use comes with tradeoffs. Data centers are often located in areas where electricity is cheap, but those same areas may face water shortages. This creates tension between the need for digital infrastructure and the need for water in communities, farming, and natural ecosystems.
Does AI Waste Water? How Much Water Is Used?
The amount of water used by data centers is substantial, but it can vary depending on the cooling system and the data center's location.
On average, a typical data center may use 300,000 to 500,000 gallons of water per day.
A large data center can use 1 to 5 million gallons of water per day, which is equal to the daily use of a small city.
In the United States, data centers are estimated to use about 1.7 billion liters of water per day.
One way experts measure water use is through the Water Usage Effectiveness (WUE) metric. This ratio compares the total water consumed to the amount of computer power delivered. A lower WUE means the data center is more efficient.
Examples from Major Companies
Several large technology companies own and operate massive data centers. Their water use has drawn attention from local governments and communities.
Google
Google has acknowledged that some of its data centers consume millions of gallons of water daily. In some cases, the company has used municipal drinking water supplies, which created tension with nearby residents.
Microsoft
Microsoft has pledged to reduce water use by developing liquid cooling systems and by recycling wastewater. However, reports show that its total water consumption rose by more than one-third in a single year because of new data center construction.
Meta (Facebook)
Meta also relies on water cooling for its servers. In some regions, its water use has sparked debates over the effect on local rivers and aquifers.
These examples show that as demand for digital services grows, water use also increases.
AI Environmental Impact
The environmental impact of water use in data centers is complex.
Local Water Shortages
In regions where water is already scarce, data center operations can put a strain on local water supplies. This may affect residents, agriculture, and wildlife.
Energy and Water Link
Water is often tied to energy use. Cooling towers, pumps, and chillers all require electricity to operate. Using more water can also mean using more power.
Wastewater
Water that passes through cooling systems may contain chemicals to prevent corrosion or bacterial growth. If not managed properly, this wastewater can harm ecosystems.
Water Scarcity Concerns
Water scarcity is becoming more severe in many parts of the world. Climate change, population growth, and farming irrigation demands all add stress to freshwater supplies. In this context, the expansion of water-intensive data centers raises difficult questions.
Should clean drinking water be used to cool servers? Can recycled or non-potable water be used as an alternative? What responsibility should companies have to the communities where they operate?
Alternatives to Heavy Water Use
There are several strategies to reduce water consumption in data centers:
Air Cooling in Cool Climates
In northern regions, outside air can be used for cooling for most of the year. This reduces the need for water-based systems.
Recycled or Non-Potable Water
Some companies are beginning to use treated wastewater from cities as an alternative to drinking water. This helps protect clean supplies.
Direct Liquid Cooling with Reuse
Advanced systems that bring cooling liquid directly to computer chips can reuse the same liquid in a closed loop, which reduces evaporation losses.
Renewable Energy and Smart Design
Placing data centers in regions with access to renewable energy and water resources can help mitigate the stress on local communities.
Community Reactions
Local communities have expressed concerns about the water use of data centers. In some towns, residents have protested new construction projects because of the potential drain on water supplies. In other cases, governments have delayed or blocked new data centers until water use agreements are reached.
This tension highlights the importance of transparency. People want to know how much water companies are using and how that use will affect their lives. Without clear communication, mistrust grows.
Balancing Technology and Sustainability
Modern society depends on digital services. However, those services have hidden costs in both energy and water. Balancing the benefits of technology with the need for environmental sustainability is one of the greatest challenges of the coming decades.
Data centers are not the only industries that use large amounts of water; however, they are growing rapidly, and the demand for their services is not slowing down. Companies, governments, and communities must work together to find solutions that allow digital progress without harming the environment.
Water plays a central role in cooling the machines that power the digital world. From search engines to online storage, every service depends on data centers, and those centers often depend on water. A single facility can consume as much water as a small city. This use affects local communities, ecosystems, and future water supplies.
At the same time, there are ways to reduce this impact. Using recycled water, enhancing cooling technology, and locating centers in cooler regions can reduce water demand. Greater transparency and responsibility from companies are also important.
The challenge is clear: we need powerful computing, and also clean water. Finding the right balance will shape not only the future of technology but also the health of communities and the environment.
Ready to transform a hidden cost into a strategic advantage? At Metis Consulting Services, we understand that sustainability isn't just a buzzword—it's a critical component of modern business, whether you're managing complex supply chains or the water footprint of your data center. We're here to help you turn environmental challenges into smart, efficient, and profitable solutions. If you're ready to stop putting out fires and start building a more resilient operation, let's chat.
Get in touch with us at hello@metisconsultingservices.com, or drop by our digital HQ at www.metisconsultingservices.com. We'll even bring the water—just for drinking, of course.
The Power of AI
Large language models (LLMs), like Gemini from Google, are emerging as powerful tools, streamlining the document creation process and allowing human expertise to shine even brighter.
This Week, the Guard Rail is thrilled to have our first-ever guest blogger. Metis' COO, Dr. Olivia Fletcher, has written a fascinating article looking deeper into AI and its use as a tool, not a replacement for human input and documentation. This comes on the heels of an exciting week at the RIC(REMS Industry Consortium) annual meeting, where our CEO, Michelleanne Bradley, presented and was on a panel discussing the intricacies of ethics and AI in the Pharmaceutical and Medical Device industries. Enjoy!
The Power of AI: How Large Language Models Are Transforming Document Creation
by Dr. Olivia Fletcher
As the COO of Metis Consulting Services, navigating a world of information and crafting clear, concise documents is essential. Traditionally, this has meant dedicating significant time to research, writing, and editing. However, the landscape is shifting. Large language models (LLMs), like Gemini from Google, are emerging as powerful tools, streamlining the document creation process and allowing human expertise to shine even brighter.
Boosting Efficiency: From Blank Page to First Draft Faster
I have ADHD, and one of my primary executive dysfunctions is task initiation. This can mean that just typing that first word is a gigantic hurdle for me. LLMs can alleviate this initial hurdle by generating drafts based on specific prompts and topics thereby providing a starting point. This can be particularly helpful for:
Emails and Reports: Quickly summarizing key points from complex data sets or research papers allows you to focus on crafting a compelling narrative. LLMs are particularly good at recognizing patterns in data.
Blog Posts and Articles: LLMs can provide a well-structured foundation, outlining the main points and even suggesting relevant sources.
This doesn't eliminate the human touch; it simply removes the initial heavy lifting.
Enhancing Content: Fact-Checking, Research, and Tone
Accuracy and credibility are paramount in any professional setting. LLMs can assist in:
Fact-checking: By integrating with vast knowledge bases, we can verify the accuracy of information and provide citations. However, we are all aware of the case of the attorney who submitted a brief in court crafted by an LLM. That brief, full of fake cases the LLM had invented for the document, is an example of the danger of just letting an LLM run away with the work. The product of an LLM still needs human verification.
Research: LLMs can efficiently scan through mountains of data and present relevant sources, saving you valuable time.
Maintaining Tone: Whether it's a formal report or a casual blog post, we can tailor the writing style to match the intended audience.
Human Expertise: Where LLMs Fall Short and We Excel
While LLMs offer significant advantages, it's crucial to remember that they are still under development. Here's where human expertise remains irreplaceable:
Critical Thinking and Analysis: LLMs can synthesize information, but they cannot replace the ability to critically analyze data, draw conclusions, and identify the underlying significance.
Creativity and Originality: Human ingenuity in crafting unique arguments, presenting information in innovative ways, and weaving a narrative is unparalleled.
Understanding Nuance and Context: LLMs may struggle with the subtle nuances of language and the importance of context in specific situations.
The Future of Document Creation: A Collaborative Approach
The ideal scenario involves a powerful synergy between LLMs and human expertise. Imagine a world where:
LLMs handle the initial groundwork: Drafting emails, reports, and even initial outlines of more complex documents.
Humans take the reins: Editing, refining the content, injecting critical thinking, and ensuring the final product aligns perfectly with the intended purpose and audience.
This collaborative approach allows professionals to:
Focus on higher-level tasks: Freeing up valuable time for strategic thinking, client interaction, and core business functions.
Produce higher quality content: The combination of LLM efficiency and human expertise produces well-structured, informative, and impactful documents.
In Conclusion: LLMs are not here to replace human writers; they are here to empower them. By embracing and utilizing this new technology strategically, professionals like myself can work smarter, not harder, and achieve even greater results.
For more information on AI and the possible thorny issues involved, listen to the Queens of Quality podcast bonus season 2.5 with guests Emily Barker and Steve Thompson S2.5(link )
To start a conversation with Metis Consulting Services, please email us at:
hello@metisconsultingservices.com
*This blog post was written with the help of Gemini, Google’s LLM.