Liquid Cooling


Why Liquid Cooling Is the Future of Data Centers


As the demand for high-performance computing (HPC) and artificial intelligence (AI) continues to skyrocket, data centers are reaching their limits in terms of power, efficiency, and operational costs. Traditional air-cooled infrastructure is no longer sufficient to support the latest generation of servers and GPUs that pack immense power into dense configurations.

According to AFCOM, the average rack density has jumped from 8.5 kW in 2023 to 12 kW in 2024, with some racks already hitting 70 kW. Expectations are growing that this could surge to 150 kW per rack in the near future. These power demands are simply too much for conventional air cooling to handle—efficiently or economically.

The industry’s answer? Liquid cooling—a solution rapidly becoming a necessity for modern data center operations.




Why Liquid Cooling?


With TDPs for CPUs and GPUs reaching 500W and 1000W respectively, the need to efficiently remove heat at the server and rack level has become critical. Liquid cooling offers a direct, highly effective method to draw heat away from components, dramatically outperforming air cooling in thermal transfer capabilities.

Unlike air cooling, which relies on airflow and large HVAC systems, liquid cooling uses water to absorb and remove heat directly from the chip surface. This closed-loop system results in quieter, more compact, and more efficient operations.




The Rise of Liquid Cooling in Data Center Design


Air conditioning cools an entire room, often inefficiently and with significant overprovisioning. In contrast, liquid cooling targets heat directly at the source—servers and components like CPUs and GPUs—making it both efficient and scalable.

The Dell’Oro Group predicts the liquid cooling market will top $15 billion in annual revenue within four years, as more data centers adopt this technology to meet growing demands.

Among liquid cooling methods, direct-to-chip cooling is the frontrunner. This technique uses cold plates attached to key components, circulating fluid to carry heat away, which is then cooled externally and recirculated. It’s effective, scalable, and compatible with both new and retrofit deployments.




The Main Reasons You Need Liquid Cooling in Your Data Center



Lower Cooling Infrastructure Costs


Air conditioning systems are large, power-hungry, and expensive to maintain. By contrast, liquid cooling requires fewer moving parts and smaller backup power supplies. Supermicro’s liquid cooling implementations, for instance, have achieved PUEs as low as 1.02, compared to 1.6 for legacy air-cooled systems—delivering up to 40% power savings.

These reductions extend beyond operations: less cooling infrastructure also means less space, less material, and lower capital expenditures during construction.


Improved Power Usage Effectiveness (PUE)


PUE is a key metric for evaluating data center efficiency. Liquid-cooled data centers come close to the ideal PUE of 1.0, dramatically outperforming traditional setups. With fewer fans, no chillers, and minimal airflow needs, cooling energy use drops significantly—up to 90% less than air cooling, according to Intel.

This lower energy footprint also reduces strain on local utilities and infrastructure, making it easier to comply with zoning and power grid restrictions.


Increased Compute Density


Liquid cooling unlocks higher compute density by eliminating the need for large heat sinks and allowing tighter server arrangements. A 4U liquid-cooled AI server can often match or exceed the performance of an 8U air-cooled system, effectively doubling computational density.

Given that data centers can cost $600 to $1,100 per square foot, this density gain translates into massive savings and more scalable performance growth.


Better Performance with Sustained Turbo Boosting


Modern processors leverage “turbo boost” technology to temporarily increase clock speeds under heavy load. But with air cooling, this boost is often short-lived as thermal thresholds are quickly reached.

Liquid cooling enables longer or even sustained boost periods, significantly improving performance for demanding workloads. Tests have shown performance gains of up to 13% for HPC applications using sustained turbo boosts under liquid cooling.


A Safer, Quieter Workplace


Air-cooled data centers can be noisy—often exceeding 96 dBA, louder than a subway train. This not only poses hearing risks but also contributes to worker fatigue and absenteeism.

Liquid cooling reduces noise by up to 55%, creating a quieter and safer environment for on-site staff. Reduced fan speed and less airflow also mean fewer maintenance points and a calmer operational atmosphere.


Lower Carbon Footprint


Liquid cooling offers a direct path to sustainability. By significantly reducing power use, it lowers carbon emissions—especially important as 82% of U.S. energy still comes from fossil fuels.

Additionally, waste heat from liquid cooling can be reused in district heating systems or other energy recovery projects, contributing to circular sustainability models.

With data center energy consumption expected to grow at 40.5% CAGR through 2027, and even faster for AI-specific facilities (up to 44.7% CAGR), adopting energy-efficient solutions is no longer optional—it’s imperative.




Supermicro Unveils DLC-2: The Future of Efficient, Liquid-Cooled Data Centers


Supermicro has introduced its next-generation Direct Liquid-Cooling solution, DLC-2, designed to drastically reduce data center energy, water usage, noise, and space requirements. By offering up to 40% power savings and cutting total cost of ownership (TCO) by as much as 20%, DLC-2 is setting a new benchmark for AI-optimized infrastructure.

The system supports warm water cooling with inlet temperatures up to 45°C, eliminating the need for traditional chillers and reducing water consumption by up to 40%. Operating at approximately 50dB, DLC-2 enables quieter environments while accelerating deployment times through a fully integrated, end-to-end solution.

The DLC-2 solution stack includes an advanced 4U server design featuring eight NVIDIA Blackwell GPUs and two Intel® Xeon® 6 CPUs, delivering high compute density with nearly full (98%) liquid-cooling coverage per rack.

The system utilizes vertical Coolant Distribution Manifolds and an in-rack CDU capable of removing 250kW of heat per rack, all managed through Supermicro’s SuperCloud Composer® software.

With support for hybrid cooling towers and a global manufacturing footprint, Supermicro offers a scalable, turnkey solution for modern data centers looking to improve efficiency, reduce environmental impact, and handle the demands of next-gen AI workloads.




Some New Supermicro Products Supported By The New DLC-2 Include:





Final Thoughts


The shift to liquid cooling is not just a trend—it’s the foundation for future-proof, AI-ready data centers. From lowering operational costs and improving performance to enabling sustainable growth, liquid cooling provides a clear and measurable advantage.

Now is the time to adopt liquid cooling—not just to stay ahead, but to stay operational in an increasingly compute-intensive world.