AI workloads push data center rack densities beyond the limits of air cooling. Three liquid cooling approaches offer different trade-offs and benefits.
August 25, 2026
As AI and high-performance computing (HPC) drive up server power densities, data centers are turning to liquid cooling to remove heat more efficiently. Rack densities have jumped from 5-10 kilowatts (kW) to 30 kW, 50 kW, or even higher – levels where air cooling struggles. More power means more heat, and liquid cooling captures it far more effectively at the source.
In the video above, tech expert Brien Posey explains why more data centers are adopting liquid cooling and what the main options look like in practice.
Comparing Air and Liquid Cooling #
Air cooling pushes cool air through servers and exhausts hot air into the aisle. This approach can work at lower heat levels but becomes increasingly inefficient at higher rack densities. Air doesn’t carry heat very well, so large volumes must be moved with bigger or more numerous fans. That adds energy costs, noise, and complexity. It also increases the risk of temperature hot spots – certain areas that get too warm, putting components under stress.
Liquids carry heat far better than air. Liquid cooling captures heat closer to the parts that produce it, often directly at the chip, so less heat escapes into the room. That means fewer or smaller fans and more stable temperatures. Many liquid systems operate with warm-water loops, which can reduce reliance on energy-intensive chillers (large machines that produce cold water) and enable the reuse of waste heat, such as for heating nearby buildings. In the right design, liquid cooling can improve performance, stability, and overall energy costs.
Three Ways Liquid Cooling Works #
The video details three options for liquid cooling:
Rear-Door Heat Exchanger (RDHx). An RDHxreplaces the rack’s rear door with a liquid-cooled coil. As servers exhaust warm air, the coil absorbs heat before it spreads into the room, helping to keep aisles cooler and reducing the load on the room’s air conditioning. RDHx is popular for retrofits because it doesn’t introduce liquid into the servers themselves and is relatively simple to install. The trade-off is that RDHx still relies on internal server fans and may be less efficient at extremely high heat loads.Direct-to-Chip Cold Plates.
Direct-to-chip coolingplaces small metal cold plates with internal coolant channels on the hottest components: CPUs, GPUs, and sometimes memory. A manifold distributes coolant to each server, while a coolant distribution unit (CDU) manages temperature, flow, and pressure and transfers heat to the facility water loop. Depending on system design, direct-to-chip can remove roughly 70-90% of a rack’s heat at the source. Fans continue to handle lower-heat components such as power supplies and networking, but overall room heat drops sharply. It’s a strong fit forGPU-dense AI training clusters.Immersion Cooling. Instead of bringing liquid to the server, immersion submerges entire servers in a non-conductive fluid. In single-phase systems, the fluid doesn’t boil; pumps move warm liquid to a heat exchanger. In two-phase systems, the fluid boils at relatively low temperatures; vapor rises, condenses on cooled surfaces, and returns to liquid. Immersion can eliminate server fans, simplify airflow management, and enable uniform cooling at very high densities. However, it requires specialized hardware, different handling and maintenance procedures, and a higher upfront cost.
Choosing the Right Approach #
Liquid cooling is a toolkit rather than a single product. RDHx can be appealing for incremental upgrades and brownfield sites. Direct-to-chip offers strong source-level heat removal for AI and HPC clusters without fully re-architecting the environment. Immersion enables very high-density deployments with uniform cooling, at the expense of more specialized infrastructure and operational costs.
The right choice depends on performance targets, energy goals, facility limitations, and total cost of ownership over the long term.
The video walks through each option and explains where it fits, so that you can weigh the trade-offs in efficiency, density, and operations.
Video Transcript #
The following transcript has been lightly edited for clarity.
Brien Posey: Hello, greetings and welcome. I’m Brien Posey, and today I want to talk a little bit about liquid cooling in the data center.
Data centers are getting hotter, literally. As AI workloads increase, so does the power that is required to run them. Modern GPU-based servers, especially those that are used for AI training and high-performance computing, are pushing rack densities far beyond what the facilities were originally designed to handle. So, whereas racks once averaged 5 to 10 kilowatts, it’s now common to see racks handle 30 kilowatts, 60 kilowatts, possibly even more.
This is where liquid cooling comes into play. And liquid cooling might sound like a cutting-edge innovation, but it’s actually been around for years. Anybody familiar with high-performance gaming PCs knows that liquid cooling has long been used to keep these systems cool. The idea is that by moving heat more efficiently than air, liquid cooling allows those systems to run faster, quieter, and more reliably under heavy loads. One way of thinking about this is that a modern AI rack is like a high-end gaming PC, but multiplied many times over.
The big challenge that liquid cooling is meant to address is heat removal. Traditional air cooling systems work by moving cold air through servers and exhausting hot air out the back. But air has its limits. It’s relatively poor at carrying heat, and moving enough air to remove all the heat requires large fans, wide paths, and a lot of energy – never mind that all those fans make a whole lot of noise.
So, at higher densities, airflow becomes harder to manage, hotspots form much more easily, and fan energy consumption rises. You’re increasing your energy costs simply by the sheer number of fans required to move enough air to keep your components cool. And cooling systems can struggle to keep up.
And this is particularly problematic as GPU counts increase, because GPUs generate heat in addition to what the server would normally produce.
To meet these challenges, data centers are turning to liquid cooling, not because it’s new, but because it’s becoming essential.
Let’s talk about why liquid cooling works better. The key advantage of using liquid cooling is simple. Liquid transfers heat way more efficiently than air does. Water or engineered coolants can carry thousands of times more heat per unit volume than air does. That means you can use a relatively small amount of liquid to remove the same amount of heat.
Heat can also be captured closer to the source, even directly from the chip that’s producing the heat, and cooling systems can operate more efficiently overall.
Another major benefit of using liquid cooling is the ability to operate at higher temperatures. Liquid cooling systems can operate at warmer coolant temperatures than traditional air-cooled systems.
And this unlocks two important advantages. First is reduced compressor use, with less reliance on energy-intensive chillers. There’s also potential for heat reuse: waste heat can be recaptured and repurposed. For example, you might use it to heat a building or for industrial processes.
In short, liquid cooling doesn’t just remove heat better; it changes the economics of cooling.
Now, the important thing to know is that there isn’t just one technology for liquid cooling. Liquid cooling in the data center spans a spectrum of technologies, each suited to different needs and use cases. We’ll talk about three technologies sometimes used to remove heat in data centers.
Rear-Door Heat Exchanger (RDHx) Cooling
The first is known as a rear-door heat exchanger, or RDHX. A rear-door heat exchanger replaces the standard back door of a server rack with a liquid-cooled coil system that removes heat from the exhaust air before it enters the room.
A rear-door heat exchanger works much like the radiator in your car. Think about a car for a moment. In a car, the engine generates a lot of heat. Coolant absorbs the heat and flows to the radiator, where air passes over the radiator coils and carries that heat away. The process keeps the car from overheating. Now, in a data center, the process is essentially flipped. Servers generate heat, like a car engine, and push hot air out the back, where it passes through a liquid-cooled coil in the rear door. That liquid absorbs the heat and carries it away. So, instead of air cooling a liquid, like what you do in a car, the liquid is cooling air. But the heat-exchange principle is the same as what you have for a car radiator.
So, how does it work step by step? Air enters the server from the cold aisle. This is the same way that it’s always worked. Then, internal server fans push that air across CPUs, GPUs, memory, and other components. Then, the hot air exits the back of the server. Instead of just going out into the room, the hot air enters the rack’s rear door, which contains a coil filled with circulating coolant. As the air passes through the coil, heat is transferred from the air to the liquid, and the air temperature drops significantly. The warmed liquid carries the heat away to the facility’s cooling system. Effectively, the rear-door heat exchangers act as a heat-capture point at the rack level.
This changes the dynamics of the entire room. The hot aisles are dramatically cooler than they would otherwise be. In some cases, particularly in low-density environments, the hot aisle might not be any warmer than the cold aisle. Also, less heat escapes into the data hall, reducing the strain on the cooling units. Airflow requirements also drop since heat is removed more effectively, and fan energy might decrease as well, especially at higher densities. In some cases, a well-designed rear-door heat exchanger can remove most of a rack’s heat load, making the room’s air-cooling system supplementary rather than the primary means of cooling the data center.
Now, when we’re talking about rear-door heat exchangers, these devices fall into two main designs, active and passive. So, a passive rear-door heat exchanger doesn’t have any fans built into the door. It relies entirely on server airflow, and it’s simple and very efficient. It’s best for moderate-density systems. Now, when we talk about active rear door heat exchangers, they include fans built into the doors designed to pull air through the coils. They can handle higher heat loads, and they provide more consistent performance regardless of server airflow.
Typically, rear-door heat exchangers are used in environments where existing systems need to be retrofitted. In other words, an organization wants to adopt liquid cooling while retaining its existing air-cooled systems and retrofitting them to use it. This is a really simple and relatively low-cost way to bring liquid cooling into the data center. It’s minimally disruptive to existing IT equipment. The operations are familiar. And there’s a relatively low risk compared to other forms of liquid cooling that bring liquid into the individual devices.
The main limitation of rear-door heat exchangers is that they still depend on air movement inside the server, and they’re also less efficient than direct-to-chip at very high rack densities.
Direct-to-Chip Cooling
With that said, let’s turn our attention to direct-to-chip cold plates. So direct-to-chip cooling delivers liquid straight to the hottest components inside a server. This might be the server’s CPU, GPU, or even memory.
The idea is that there’s a liquid-cooled cold plate that sits in direct contact with those components. And this is very, very similar to the methods used in high-end gaming PCs.
If you’ve got a PC like that with a liquid cooler, then you’ve already seen a version of this. In a gaming PC, there’s a metal block. It’s called a water block that sits directly on the CPU or GPU. Liquid flows through that block, absorbing heat from the component. Then, that warmed liquid travels to a radiator. Fans pass air through the radiator, where the heat is released. So, direct-to-chip cooling in a data center works on a very similar principle. It’s just scaled up and engineered for reliability, serviceability, and continuous operations.
How does this work step by step? The idea is that cold plates are mounted directly onto CPUs, GPUs, or other hot components, and then coolant flows into the rack through a facility connection. It enters a manifold that distributes the liquid to each server. Then, there’s a quick disconnect that allows servers to be serviced without draining the system. Liquid flows through each cold plate, absorbing heat directly from the chip’s surface. Then that liquid exits the server and returns to the rack loop. The liquid is routed to a coolant distribution unit (CDU), which transfers heat to the facility water system while isolating the IT coolant loop and regulating temperature flow and pressure.
So, direct-to-chip is very effective. It greatly reduces the amount of heat released in the room. It allows for much higher rack densities without overwhelming air systems.
Fans still exist, but they do something a little bit different. Typically, fans are used to air-cool components under lower load. This might be things like power supplies or networking components, or just the residual heat coming off the server. But then liquid cooling is used for the hotter components: the CPUs, GPUs, memory, and the like. But the majority of the heat, often 70 to 90% of the heat, is captured for liquid. And this works really well for AI training clusters, GPU-dense environments, and high-performance computing.
Immersion Cooling
Now, the third type of liquid cooling I want to discuss is immersion cooling, which is fundamentally different from the other two forms I’ve discussed. Instead of bringing cooling to the server, it literally submerges the entire server in the liquid. Yes, the server is submerged in liquid.
Now, we’re not talking about water here. We’re talking about a special liquid designed to be dielectric, meaning it doesn’t conduct electricity. This means your server components can operate safely while submerged in liquid.
There are two main types of immersion cooling. There are single-phase and two-phase. Single immersion is like putting a hot pan into a sink full of water. The water absorbs heat and warms up gradually, but the pan cools very, very quickly as soon as it’s submerged in the water. Two-phase immersion is like boiling water on the stove. Heat causes the liquid in the pan to boil, which is a “phase change.” As the boiling process occurs, it carries heat away from the pan.
Let’s talk more about single-phase and two-phase immersion. With single-phase immersion, servers are submerged in a dielectric fluid bath. This fluid doesn’t boil. Instead, pumps circulate the warm fluid through the heat exchanger, where the heat is transferred to the facility water and carried away. So, what’s happening physically is that the heat moves from the component to the surrounding liquid and then ultimately into a cooling loop.
Two-phase immersion works very differently. What happens is that the servers are still submerged in liquid, but this liquid is designed to boil at very low temperatures, so the heat from the chips causes it to boil, and the vaporized liquid evaporates. It turns into a steam, and then the steam comes into contact with a cooling plate that’s just above the tank full of liquid. And when that steam comes into contact with the cooling surface, it condenses back into a liquid. Then, the liquid returns to the cooling tank. And so, that’s what causes the heat to be removed. The heat is being removed by a phase change. In other words, that conversion from liquid to steam and back into liquid.
The big advantages of this type of liquid cooling are minimal need for server fans, lower energy use, no need to move large volumes of air, uniform cooling across all of the components, and extremely efficient heat transfer. This means you can achieve much higher server densities. It also reduces acoustic noise because you don’t need all those fans, and it simplifies airflow management.
However, there are some limitations. For one thing, it requires an entirely new operational model. You can’t just dunk your existing rack-based systems in a tank of liquid. It doesn’t work that way. You’re going to need specialized hardware. So, there will be hardware compatibility and handling considerations. Also, you have to think about fluid management, long-term maintenance, and planning, because remember, your servers are submerged in liquid, so they’re not quite as easily accessible as they would have been before. There’s also a higher initial implementation cost, so that’s something to consider.
And just a side note: immersion cooling is no longer just for cryptocurrency mining. While it gained early attention in that space, this type of cooling is now being piloted, tested, and increasingly adopted for enterprise, AI, and high-performance computing workloads, where density and efficiency are becoming critical constraints.
A Range of Options
The big picture is that liquid cooling isn’t a single solution; it’s a range of options.
Rear-door heat exchangers extend the life of air-cooled environments and are comparatively inexpensive to implement. Direct-to-chip systems target the most intense heat sources. Immersion rethinks cooling entirely to achieve maximum density and efficiency.
The important thing to keep in mind is that liquid cooling is a spectrum, not a single technology, and it’s becoming essential for the next generation of data centers.