Cooling methods for data centres:
liquid cooling vs immersion cooling

Why efficient cooling in the data centre is becoming increasingly important

For years, the go-to response to rising heat was to install more powerful fans and turn up the air conditioning. But this approach is now hitting technical and economic limits. With the increasing power density of modern servers and the heat generated by specialised GPUs, traditional air cooling is simply no longer up to the task – it’s inefficient, power-hungry, and takes up too much space. Today’s data centres need smart solutions that keep performance and energy consumption equally under control.

The insatiable energy appetite of AI and HPC

Processors for high-performance computing (HPC) and modern AI accelerators are marvels of engineering – and at the same time, serious heat generators. Just a few years ago, the thermal design power (TDP) of a typical server GPU was between 150 and 300 watts. Today, a single Nvidia Blackwell B200 Tensor Core reaches up to 1,200 watts. In systems like the GB200, these high-performance GPUs are even used in pairs.

As a result, not only do performance figures skyrocket – the demands on thermal management rise dramatically too. Traditional air cooling quickly becomes a bottleneck and a cost driver. It can account for up to 40% of a data centre’s total energy consumption, which results in poor PUE values (often above 1.4) and soaring electricity bills. In times of strict ESG requirements, this level of inefficiency is simply no longer acceptable.

Air cooling at its limit: Chipmakers pave the way for new methods

The major chip manufacturers are actively driving this transition:

• Nvidia has designed its new Blackwell architecture around the GB200 superchip specifically for direct liquid cooling. Only liquid can reliably tame such extreme power densities.
• Intel and AMD are working closely with technology partners to validate and optimise their most powerful CPU families – Xeon and EPYC – as well as their GPU accelerators for liquid cooling.

The message is clear: if you want to unlock the full potential of the next hardware generation, liquid cooling is no longer optional.

Liquid cooling: proven technology with new possibilities

Liquid cooling uses a coolant to dissipate the heat generated during operation. A particularly efficient method is so-called direct-to-chip cooling, where the coolant is directed straight through specialised cold plates on processors, GPUs, or other hotspots. This approach is like a precise surgical intervention: instead of cooling the entire room or server, the cooling is applied exactly where the most heat is produced.
Liquid cooling technology is no longer a niche solution – in recent years, it has established itself as a powerful alternative to traditional air cooling.

The advantages of liquid cooling

• Targeted cooling for your power chips: Liquid conducts heat 25 times better than air. That means even heat monsters drawing over 1,000 watts are kept under control – allowing your AI and HPC systems to operate at full throttle.
• More computing power per rack: With no need for bulky fans or wide air corridors, you can pack your racks more densely and fit more compute power into the same footprint.
• Peace and quiet in the server room: Fewer fans means much less noise. That’s good news for your ears – and your colleagues’.
• Heating instead of just cooling: The waste heat generated by modern DLC systems becomes a valuable resource. Use it to heat your building and reduce operational costs – a major plus for your ESG balance sheet.
• Perfect for retrofitting: Thanks to their modular design, many DLC systems can be easily integrated into existing 19-inch racks. Ideal for modernising your data centre infrastructure.

Liquid cooling in practice: Use cases and applications

A prime example is Nvidia: the new GB200 NVL72 system was designed from the ground up for direct liquid cooling in order to keep its immense performance in check. Leading hyperscalers are embracing this trend, and even enterprise data centres are adopting DLC for performance-intensive workloads.

Typical use cases include:

• High-Performance Computing (HPC): Enables maximum compute power in research and science without thermal throttling during complex simulations.
• Artificial Intelligence (AI) & Machine Learning (ML): Maximises performance per watt during large model training and shortens compute times.
• Financial services: Delivers top performance in tight spaces – essential for latency-sensitive high-frequency trading.
• Media & Entertainment: Ensures stable and quiet continuous operation during demanding 3D rendering tasks.

Limitations & challenges of liquid cooling

As efficient as liquid cooling may be, it does come with some challenges. Installing pumps, cold plates and coolant loops is more complex and requires specialist know-how. Water-based systems must be regularly checked for leaks – even minor leakage can have serious consequences for hardware.

And keep in mind: direct liquid cooling (DLC) only cools the “stars” – like CPUs and GPUs. Other components such as memory or network chips still rely on air cooling. That means you’re managing a hybrid system and must ensure there are no thermal imbalances between different cooling zones. Expert knowledge is essential.

Immersion cooling: a quantum leap in cooling architecture

Why settle for less when you can go all-in? Immersion Cooling submerges your entire server in a non-conductive liquid bath. The coolant envelops every component, dissipating heat silently and evenly – without any fans. There are two methods:

• Single-phase immersion: The coolant stays liquid, flows over the components, absorbs heat, and is cooled externally in a simple loop.
• Two-phase immersion: Things get physical here. The coolant boils at hot spots, rises as vapor, condenses at the top, and rains back down onto the hardware.

The benefits of immersion cooling

• Maximum cooling power for extreme densities: Every component is bathed in coolant, eliminating hot spots and delivering unmatched thermal stability – even at the highest compute densities.
• Massive energy savings: Without fans or air conditioning, you can cut cooling energy use by up to 95%. Your PUE drops significantly – and room temperature becomes irrelevant.
• More power in less space: Forget about hot-aisle/cold-aisle separation. Immersion tanks sit close together, enabling extremely high compute density and better space efficiency.
• Longer hardware life & better reliability: Immersed hardware is protected from dust, moisture, and vibration. Stable temperatures reduce component wear and increase overall system reliability.
• Efficient heat reuse & sustainability: Coolants in immersion systems reach higher temperatures, making heat recovery more effective. Reuse the waste heat for building heating or feed it into district networks – great for cutting costs, reducing CO₂ emissions, and meeting ESG and green IT goals.

Immersion cooling in practice: use cases and real-world examples

Immersion cooling has grown from a niche concept into a field-proven solution for ultra-dense workloads. A standout example is RNT’s iXora HRM-4: instead of open tanks, the system uses closed cooling cassettes that fit directly into 19-inch racks. Each cassette handles up to 4 kW of thermal load (up to 16 kW per chassis), making it ideal for even the most demanding AI and HPC workloads.

Thanks to its high power density, immersion cooling is proving valuable in several high-demand environments:

• High-Performance Computing (HPC) & Artificial Intelligence (AI): Enables dense, thermally stable clusters with no throttling.
• Blockchain & Crypto Mining: Provides cost-effective, reliable cooling for power-hungry rigs.
• Edge Computing: Creates compact, ruggedized systems for space-constrained or harsh environments.
• R&D Labs: Allows experimental hardware and prototypes to run under perfect thermal conditions.

Conclusion: the right cooling strategy for your data centre

Data centres are the backbone of the digital world — but they need the right environment to operate efficiently and reliably. Immersion Cooling and Liquid Cooling offer two innovative solutions to the thermal challenges of modern IT. Which one is better? That depends entirely on your specific requirements.

One thing is clear: those who set the direction for investment today will determine the performance, operating costs, and ESG footprint for years to come. And often, the return on investment comes sooner than expected. At RNT, we’re here to help you make the right choice — with the right technology, expert advice, and experience gained from countless projects.

Sources used