Artificial intelligence (AI) has raised the stakes in every area of technology, including one of the most traditional: data centers. Modern data centers are running hotter than ever—not only to handle ever-increasing processing demands, but also because of rising temperatures as a result of AI workloads, with no end in sight. Two in three IT leaders expect spending on AI to increase this year, requiring new ways to cool data centers.(1)
Liquid cooling is rapidly gaining importance because it is the future of data center success. A survey in spring 2024 of 812 IT professionals(2) found that 38.3% expect to deploy liquid cooling infrastructure in their data centers by 2026. That’s nearly double the percentage of a similar survey conducted in early 2024, and this accelerating trend is likely to continue over the next few years.
An indispensable function
Liquid cooling is a simple concept: A circulating liquid removes excess heat from the unit. More specifically, direct-to-chip liquid cooling uses water as the primary coolant, while other cooling techniques – such as immersion cooling and traditional air cooling – have limited capacities. Future-proof data centers have used liquid cooling for decades, but the latest chips – which can reach temperatures of up to 2,000 watts, about 20 times what basic computer chips conduct – require new solutions. Intel, NVIDIA and other AI chip makers are focused on increasing processing power, which in turn increases the amount of heat released. Coupled with industry demand for the chips, liquid cooling solutions not only make sense, but are a necessity for efficient, sustainable and reliable data centers.
Liquid cooling is an essential feature in the data center for three main reasons. These include:
- Coping with the increasing power density of modern data centers
- Room for the newest and hottest chips
- Reduce energy costs and achieve greater sustainability
The continuous nature of AI workloads means they typically cannot pause mid-query or save their space during an outage. “Unexpected downtime due to overheating would be catastrophic,” says Gary Rowe, executive chairman at Chilldyne. “The potential for hardware damage from coolant leaks is also enormous. With individual AI chipsets costing over $30,000 each, unreliable liquid cooling methods are both impractical and prohibitively expensive.”
Implementation of liquid cooling
These requirements also make it important to have the right partner for a liquid cooling solution. Chilldyne, for example, is an innovator in the liquid cooling space as the company uses a proprietary, spill-proof direct-to-chip system called negative pressure technology.
Liquid cooling can dissipate heat 1000 times better than air systems, but liquid cooling with negative pressure draws the coolant through the system instead of pushing it. This reduces the risk of water leaks and protects chipsets from overheating.
Chilldyne’s solutions can support both AI and traditional high-performance computing (HPC) chips and handle heat loads of 2,000 watts and more per single processor. In addition, the company has expertise in building new AI-driven data centers as well as retrofitting existing data centers to accommodate liquid cooling and provide “unprecedented reliability, efficiency and sustainability,” Rowe said.
The technology is so important that the company recently received a grant from the U.S. Department of Energy’s Advanced Research Projects Agency-Energy (ARPA-E) to support ongoing development of energy-efficient cooling solutions. Chilldyne’s 2,000-watt rated turbulator cold plates with automated coolant quality (ACQ) are receiving special funding from the Department of Energy.
Leadership means building a foundation that not only supports today’s capabilities but also meets the technical requirements of the next generation. As the demands on AI systems rapidly increase, forward-thinking data center managers are now turning to liquid cooling.
For more information, visit Chilldyne.
(1) Foundry, AI Priorities 2023 Study, https://foundryco.com/tools-for-marketers/research-ai-priorities/
(2) The Register, Spring 2024 survey, https://www.theregister.com/2024/04/22/register_liquid_cooling_survey/
