October 23, 2026
This is the fifth in a series of general education articles of relevant topics that we think you may find of interest. If nothing else, this insight may help you look and sound more knowledgeable with your golf foursome, at pickleball, or at your local watering hole.
AI data centers use a lot of energy and water for one basic reason: artificial intelligence takes an enormous amount of computing power, and computing power creates heat. The more powerful the AI system, the more electricity it needs. The more electricity it uses, the more heat it produces. That heat must be removed so the equipment does not slow down, malfunction, or fail.
A simple way to think about it is this: an AI data center is not just a room full of computers sitting quietly. It is more like an industrial plant filled with thousands of powerful machines working at the same time, often all day and all night. Those machines are doing the math that allows AI tools to answer questions, write text, analyze documents, generate images, recognize patterns, and make predictions.
The biggest reason AI data centers use so much electricity is the chips. Modern AI systems rely heavily on specialized computer chips called GPUs, along with other AI accelerators. These chips are different from the ordinary processors found in most laptops or office computers. They are designed to do many calculations at once, very quickly. That makes them excellent for AI, but it also means they use much more power.
One high-end AI chip can use as much electricity as several household appliances running at the same time. A single data center may contain thousands, or even tens of thousands, of these chips. When they are linked together to train or run large AI models, the total power demand becomes enormous.
AI also tends to run continuously. A home computer may sit idle most of the day. A regular company server may have busy periods and quiet periods. But an AI data center is often designed to keep its chips working close to full capacity for long stretches of time. Training a large AI model can take weeks or months. After that, running the model for users can also require constant work, especially when millions of people are using AI tools at the same time.
This user-facing part of AI is called inference. Every time someone asks a chatbot a question, asks an AI program to summarize a file, or generates an image, the data center must perform calculations to create the answer. One request may not seem like much. But billions of requests across many users, businesses, and products can add up to a major electricity load.
There is another energy cost that is easy to miss: moving information around. AI chips need to communicate with each other very quickly. Data moves between chips, memory, storage systems, and networking equipment. That movement also uses electricity. In very large AI systems, the supporting equipment can become a meaningful part of the total power bill.
Nearly all the electricity used by computer equipment eventually turns into heat. This is true for a laptop, a phone, and a data center. The difference is scale. A laptop might get warm on your desk. An AI data center can produce heat on the scale of a large industrial facility.
If the heat is not removed, the chips cannot operate safely or efficiently. They may slow themselves down to avoid damage, which reduces performance. In more serious cases, overheating can shorten equipment life or cause failures. For this reason, cooling is not optional. It is a core part of the data center’s operation.
Cooling also uses additional electricity. Fans, pumps, chillers, cooling towers, and other systems all require power. So the data center uses electricity first to run the computers and then uses more electricity to keep those computers cool enough to operate.
Water matters because it is incredibly good at absorbing and carrying away heat. Anyone who has touched a hot pan with a wet cloth or felt cooler after sweating has experienced the same basic idea. Water can capture heat efficiently, and evaporation can remove a great deal of heat with relatively little electricity.
Many data centers use water-based cooling systems because they can be more energy-efficient than relying only on traditional air conditioning. In some cases, water is circulated through pipes or cooling equipment to carry heat away from servers. In other cases, water is allowed to evaporate as part of the cooling process. When water evaporates, it takes heat with it. That helps cool the facility, but it also means some of the water is consumed and cannot simply be reused on site.
This creates a trade-off. A cooling system that uses more water may use less electricity. A cooling system that uses less water may need more electricity to run chillers, fans, or other equipment. There is no single perfect solution. The best design depends on the location, climate, electricity supply, water availability, and type of data center.
Water use can vary widely. A data center in a cool climate may need less cooling than one in a hot region. A facility using advanced closed-loop liquid cooling may consume less water than one relying heavily on evaporative cooling. Some operators also use reclaimed or non-drinking water to reduce pressure on local drinking-water supplies.
There is also indirect water use. Some of the water footprint of an AI data center comes from the power plants that generate its electricity. Many power plants, including natural gas, coal, and nuclear plants, use water for cooling. That means a data center may use water both directly at the facility and indirectly through the electricity it buys from the grid.
This is why the source of electricity matters. A data center powered by electricity from sources that require less water will have a different overall footprint than one powered mainly by thermal power plants. The environmental impact depends not only on how much electricity is used, but also on how that electricity is produced.
Data center operators are trying to reduce both energy and water use. New chips are becoming more efficient, meaning they can do more AI work for each unit of electricity. Software developers are also working on smaller and more efficient AI models that require less computing power for certain tasks.
Cooling technology is also improving. Direct-to-chip liquid cooling brings cooling fluid closer to the hottest parts of the equipment, which can remove heat more efficiently than air alone. Closed-loop systems can reuse fluid rather than constantly consuming new water. Some facilities are also exploring immersion cooling, where equipment is placed in special non-conductive liquid that absorbs heat directly.
Companies are also paying more attention to where they build. A data center located near reliable power, cooler weather, or available recycled water may have fewer environmental challenges than one built in a hot or water-stressed area. In some cases, operators are looking for dedicated power supplies, including renewable energy, nuclear power, or other sources, to reduce pressure on local electric grids.
AI data centers use a lot of energy because AI requires thousands of powerful chips working for long periods of time. They use water because cooling those chips is difficult, and water is one of the most effective ways to move heat away from the equipment. The key issue is not just that AI uses resources. The key issue is scale: as demand for AI grows, the physical infrastructure behind it must grow too. The future will depend on whether the industry can keep improving efficiency fast enough to offset rising demand. Better chips, smarter software, cleaner power, more efficient cooling, and careful site selection can all help. But the basic physics will remain the same: computing uses electricity, electricity creates heat, and heat must be removed. That is why energy and water are now central issues in the growth of artificial intelligence.
Landsberg Bennett is a group comprised of investment professionals registered with Hightower Advisors, LLC, an SEC registered investment adviser. Some investment professionals may also be registered with Hightower Securities, LLC (member FINRA and SIPC). Advisory services are offered through Hightower Advisors, LLC. Securities are offered through Hightower Securities, LLC.
This is not an offer to buy or sell securities, nor should anything contained herein be construed as a recommendation or advice of any kind. Consult with an appropriately credentialed professional before making any financial, investment, tax or legal decision. No investment process is free of risk, and there is no guarantee that any investment process or investment opportunities will be profitable or suitable for all investors. Past performance is neither indicative nor a guarantee of future results. You cannot invest directly in an index.
These materials were created for informational purposes only; the opinions and positions stated are those of the author(s) and are not necessarily the official opinion or position of Hightower Advisors, LLC or its affiliates (“Hightower”). Any examples used are for illustrative purposes only and based on generic assumptions. All data or other information referenced is from sources believed to be reliable but not independently verified. Information provided is as of the date referenced and is subject to change without notice. Hightower assumes no liability for any action made or taken in reliance on or relating in any way to this information. Hightower makes no representations or warranties, express or implied, as to the accuracy or completeness of the information, for statements or errors or omissions, or results obtained from the use of this information. References to any person, organization, or the inclusion of external hyperlinks does not constitute endorsement (or guarantee of accuracy or safety) by Hightower of any such person, organization or linked website or the information, products or services contained therein.
Click here for definitions of and disclosures specific to commonly used terms.