A look inside closed-loop data center cooling, how it differs from traditional cooling towers, and why AI is accelerating the shift.
Over the last few years, data center water use has gone from a fairly technical infrastructure issue to a very public one. Communities have pushed back on large projects, headlines have focused on the amount of water needed to cool AI infrastructure, and technology companies have faced increasingly difficult questions about whether rapidly expanding computing capacity should also mean rapidly expanding demand on local water supplies.
While some of that discussion may have been oversimplified, much of it has been completely fair. Either way, the pressure has mattered.
Data center operators were already looking for ways to cool equipment more efficiently, but the growth of AI has made the problem much harder to ignore. More computing power means more heat, and simply scaling up older cooling approaches can mean more demand on water systems that may already be under stress. That has helped push cooling technology in a different direction, where the goal is increasingly not just to cool more efficiently, but to do it with less ongoing water consumption.
One of the biggest changes is the growing use of closed-loop cooling. The idea itself is not new, but the way it is being applied inside modern data centers is changing quickly. To understand why that matters, it helps to start with the difference between closed-loop cooling and the evaporative systems that helped give data centers their reputation for heavy water use in the first place.
Closed-Loop Cooling vs. Traditional Evaporative Cooling
Traditional evaporative cooling and closed-loop cooling can both use water. The difference is what happens to that water after it picks up heat.
In an evaporative system, water is intentionally exposed to air so that some of it evaporates. That evaporation carries heat away, which is why cooling towers have been used for decades in power plants, factories, office buildings and data centers. It is a very effective way to cool a large amount of equipment, but it also uses water by design.
Once part of that water evaporates into the atmosphere, it has to be replaced. Because the minerals and treatment chemicals in the water do not evaporate with it, they also become more concentrated over time. Eventually, some of the circulating water has to be removed as cooling tower blowdown and replaced with fresh water.
That is an important part of the water-use story around traditional data centers. The system is not simply circulating the same water forever. Some of it is continually being lost through evaporation, and some of it eventually leaves as wastewater.
Closed-loop cooling works differently. The cooling fluid stays inside the system, absorbing heat, carrying it away, releasing that heat somewhere else and then returning to do the same thing again. That difference matters quite a bit when you start talking about water use. An evaporative system is designed around losing some water as part of the cooling process, while a closed loop is designed around keeping that fluid in circulation.
That does not mean closed-loop systems are automatically water-free, which we will get to in a moment. But it is a fundamentally different approach.
How a Closed-Loop Cooling System Actually Works
The easiest comparison may be the cooling system in a car. Coolant moves through the engine, absorbs heat and then travels through the radiator, where that heat is released. The coolant then returns to the engine and repeats the process. You do not keep pouring fresh coolant into the car while you drive because the same fluid stays inside the system and keeps circulating.
A data center cooling loop follows the same basic idea, although the equipment is much more sophisticated. The cooling fluid may be treated water, a water-and-glycol mixture or another engineered coolant. It moves through piping, picks up heat from the computing equipment, carries that heat to another part of the system and then returns to do it again.
The same fluid can make that trip over and over, and that is the part that changes the water equation. Instead of constantly replacing cooling water because it has evaporated, the system is designed to reuse the same fluid for long periods of time. Where things get more interesting is in how close that liquid can get to the computers themselves.
Instead of Chasing Heat Around the Room, Move the Cooling Closer to the Source
Traditional data centers have relied heavily on air. Servers generate heat, fans move that heat into the room, cooling equipment captures it, and cooler air is sent back toward the servers. That approach has worked for a long time, but AI is making it harder.
Modern AI hardware can produce a tremendous amount of heat in a very small space. At a certain point, moving more and more air around the room becomes a less attractive way to solve the problem, so the industry is increasingly trying to stop chasing heat around the room and start collecting it closer to where it is being created.
One of the main approaches is direct-to-chip liquid cooling. A cold plate sits directly against high-heat components such as CPUs and GPUs, and cooling liquid moves through small channels in that plate. Instead of waiting for the processor to heat the surrounding air first, the liquid begins removing heat almost immediately.
In simple terms, the cooling system gets much closer to the source of the problem. That becomes especially useful in dense AI environments, where individual racks can generate much more heat than the server rooms of the past were designed to handle. It is one reason closed-loop liquid cooling has gone from something relatively specialized to something now being discussed as a major part of future data center design.
Closed Loop Does Not Automatically Mean Zero Water Use
This is where the terminology can get confusing. A data center can have a completely closed cooling loop around its servers and still have a cooling tower outside using water.
The internal liquid may collect heat from the computers and transfer it through a heat exchanger into another cooling system. If that second system eventually gets rid of the heat through an evaporative cooling tower, the facility is still consuming water even though the server-side loop itself is closed.
That is why simply hearing that a data center uses "closed-loop cooling" does not tell you the whole story. The bigger question is how the facility ultimately gets rid of the heat.
If it uses a cooling tower, some water may still be lost through evaporation. If it uses a dry cooler or air-cooled chiller, it may be able to release heat into the outside air without continually consuming water. Some facilities may use a combination of both approaches depending on weather, energy demand and cooling needs.
So closed loop is not really the end of the water-use discussion. It is the beginning of a more useful one.
AI Is Making This Change More Urgent
The public conversation around data center water use did not create the need for better cooling by itself. AI hardware was already pushing the industry toward new solutions because the heat loads were becoming more difficult to manage, but public scrutiny has added another reason to move faster.
If a company is building a large new data center in a community already worried about drought, municipal water capacity or competing demand, cooling technology suddenly becomes more than an engineering decision. It becomes part of the public case for whether that project fits in that location.
That helps explain why major technology companies are investing so heavily in systems that reduce evaporative water use, rely more heavily on closed loops or use non-potable and reclaimed water where evaporation is still part of the cooling strategy. The water question has become part of the design conversation much earlier than it used to be.
What This Looks Like in New Data Centers
Some of the clearest examples are already appearing in newly announced and recently built facilities. Microsoft, for example, has discussed newer data center designs that use closed-loop liquid cooling systems intended to circulate cooling water rather than continually consume it through evaporation during normal operation.
Other operators are experimenting with combinations of direct-to-chip liquid cooling, dry coolers, air-cooled chillers and hybrid systems that can change how they reject heat depending on local conditions. There is not one universal new cooling system replacing the old one. What seems to be emerging instead is a much wider range of options.
A data center in a hot, water-stressed region may make very different cooling choices than one built in a cooler climate with abundant reclaimed water. A facility designed around high-density AI computing may also look very different from a more traditional enterprise data center. As a result, two facilities with roughly similar computing capacity could eventually have very different water footprints.
And Then There Is Immersion Cooling
Direct-to-chip cooling brings liquid very close to the processor. Immersion cooling goes further.
In an immersion system, servers or computing components are placed directly into a specially engineered dielectric fluid that does not conduct electricity. Heat transfers directly from the electronics into the fluid instead of first moving through the surrounding air. It sounds futuristic, but it is a real technology that has been under development and deployment for years.
That does not necessarily mean every future data center will be filled with servers sitting in tanks of fluid. More likely, immersion cooling will become another tool alongside direct-to-chip cooling, air cooling, closed-loop water systems and hybrid approaches.
The future of data center cooling is probably not one technology replacing everything else. It is more likely to be a much more complicated mix.
The Future Is Probably Not "Water" Versus "No Water"
This may be the biggest change in how data center water use should be discussed. For years, the conversation was relatively simple: servers create heat, cooling towers remove that heat, and cooling towers consume water. That model is still very real, but it is just no longer the only model.
Newer facilities may use closed-loop liquid cooling around their highest-density equipment, dry cooling when weather conditions allow it and evaporative cooling only when additional capacity is needed. Other sites may use reclaimed water instead of drinking water, while some may be designed to avoid routine evaporative cooling almost entirely.
Every approach comes with tradeoffs. Reducing water consumption can increase the importance of electricity use. Evaporative cooling can be extremely efficient in the right climate, while dry cooling may perform differently during very hot weather. Local water availability can completely change which option makes the most sense.
There is no completely consequence-free way to remove an enormous amount of heat, but there are now more ways to do it than there were even a few years ago.
Closed Loops Still Eventually Create Wastewater
There is one other part of the closed-loop story that tends to get overlooked: the fluid may stay in the system for a long time, but it does not stay there forever.
Before a new cooling system begins operating, its piping has to be cleaned and flushed. Construction debris, suspended solids, metals, oils, corrosion products and treatment chemicals can all be present during those initial flushes. The cooling loop itself may also contain corrosion inhibitors, glycol and other additives once it enters normal service.
Later, maintenance, repairs, chemistry changes or equipment replacement can require some or all of that fluid to be removed. So while the wastewater does not disappear, the pattern changes.
A traditional evaporative system may continually use makeup water and regularly generate blowdown. A closed-loop system may use far less water during normal operation but generate significant wastewater during commissioning, maintenance or system changeouts. The water is still part of the conversation. It simply shows up differently.
Public Pressure May End Up Changing More Than the Headlines
The criticism surrounding data center water use has not always been perfectly precise, but it has focused attention on a real issue, and that attention is helping push the industry toward better answers.
AI is creating more heat than ever, but it is also arriving at a moment when communities, regulators and technology companies are paying much closer attention to where cooling water comes from, how much is consumed and what happens to it afterward. That does not mean future data centers will stop using water, but it does mean the phrase "data center water use" is going to become less useful unless we also ask how that particular facility is being cooled.
A better question may be:
What kind of cooling system is being used, and where does the heat ultimately go?
The answer to that tells us much more about how much water the facility may actually need.
Want to Learn More About Data Center Water?
Data center cooling is only one part of the water and wastewater picture. Valicor has additional resources covering commissioning flushwater, glycol and antifreeze, cooling tower blowdown, and other wastewater streams generated during construction and operation.
Explore more of our Data Center Wastewater Resources: