Stay Cool, Compute Faster

In modern data centers, the power consumption of servers continues to rise — driven primarily by compute-intensive applications such as Artificial Intelligence (AI) and High-Performance Computing (HPC). The processors and chips in these systems now reach thermal power losses of several hundred watts per component. Traditional air cooling is reaching its limits, as it can no longer dissipate the generated heat efficiently without consuming a disproportionate amount of energy for air conditioning.
Liquid cooling, also known as liquid-based or water cooling, offers an effective alternative. It enables significantly higher heat dissipation capacity while reducing the energy required for cooling.

There are several approaches to using liquids for heat removal in data centers:

Rack-Level Liquid Cooling – Cooling plates or heat exchangers are integrated directly into the rack to absorb heat from the air or from heat sinks.
Rear Door Heat Exchanger – Water-cooled doors at the back of the rack cool the warm exhaust air before it enters the room.
Immersion Cooling – Servers are completely immersed in a non-conductive dielectric liquid.
On-Chip Cooling – Cooling channels are integrated directly into the chip or substrate, allowing coolant to flow right past the hotspots.
Direct-to-Chip Cooling – Liquid is circulated through cooling plates that sit directly on processors, GPUs, or other components.

At the Heart of the Processor

In Direct-to-Chip Cooling, the coolant doesn’t just pass over surface-mounted heat sinks — it is guided directly over specially designed cooling plates attached to the top of the processors. These plates feature extremely fine microchannels through which the coolant flows directly past the thermal hotspots of the chips — precisely where the computing cores generate the most heat.
Because the coolant flows so close to the heat source, the heat has to travel only very short distances. This minimizes temperature differences and greatly increases cooling efficiency.

How It Works

The coolant circulates in a closed loop: it absorbs the waste heat from the chip hotspots in the cooling plates and transports it to an external heat exchanger. There, the heat is transferred to a secondary cooling system, such as a central chilled-water supply.
Since several thermal transition layers — such as thermal pastes or additional heat spreaders — are eliminated, the overall thermal resistance is significantly reduced. This results in more efficient cooling and enhanced overall system performance.

Advantages of Liquid Cooling

The advantages of Direct-to-Chip Cooling lie primarily in its maximum proximity to the heat source. Because heat is removed directly at the chip, surrounding components hardly heat up at all. This allows computing units to be packed more densely and enables chips to run at higher clock speeds or with more cores — without exceeding thermal limits.
At the same time, the energy demand for data center air conditioning is reduced, improving overall energy efficiency and positively affecting the PUE (Power Usage Effectiveness).
An example of integrating this technology is the “Data Center in a Box” — the DC IT Container from the Data Center Group (DCG). Container-based solutions make it possible to deploy high-performance infrastructures within a few months — tailored to both current and future hardware generations. Despite rising demands on power, cooling, and infrastructure, the solution remains compact, flexible, and efficient.

Challenges

However, the technology also poses challenges for design and implementation. Manufacturing cooling plates with fine microchannels requires high-precision processes, and because the coolant flows directly past the chip, absolutely reliable sealing concepts are essential to prevent leaks.
Component replacement can also be more complex, as the cooling plates are an integral part of the server architecture. Nevertheless, the benefits clearly outweigh the drawbacks: Direct-to-Chip Cooling is no longer limited to high-performance data centers or exascale supercomputers — it is increasingly relevant for modern AI infrastructures, where thousands of GPUs run in parallel, as well as for specialized systems in fields like cryptography or blockchain.
For Direct-to-Chip Cooling to reach its full potential, it is crucial that IT components and infrastructure are planned together. Only when power supply, cooling systems, rack design, and hardware are conceived as a single, integrated system can safe, efficient, and future-ready data center environments be created.
The experts at DCG are ready to assist you with all these aspects.

Looking Ahead

Direct-to-Chip Cooling is regarded as one of the most promising future technologies to handle the growing thermal challenges of modern processors — making it a key factor for the next generation of high-performance data centers.
We are ready for this future. DCG designs highly available data centers that enable the use of innovative cooling technologies — working closely with strong partners.
Get in touch with us — together, we’ll shape the data centers of tomorrow!

Title picture © stone36 / #193727940 / stock.adobe.com (Standard licence)