An Additional Look At Liquid Cooling

I previously wrote about the potential of liquid cooling for high power LED luminaires, here. Liquid cooling is becoming increasingly popular for AI servers in advanced data centers, the charging cables for high-powered Level 3 EV chargers, and even next gen EV batteries. Traditional air-cooling methods can no longer meet thermal dissipation requirements for these high-powered applications. Liquid cooling solutions, including immersion cooling, have emerged as a new solution, particularly with Nvidia’s advocacy for AI servers.

Advanced technologies like cloud computing, machine learning, and AI are making traditional air cooling methods inadequate. Data center revenues are expected to grow at double-digit rates, while liquid cooling’s revenues are projected to increase by 17% in 2024 and more than 30% annual growth after 2027.

Single-phase cold plates currently hold a higher market share for servers, compared to immersion cooling (3:1 ratio). However, as CPUs and GPUs evolve, stricter thermal management requirements and improved Power Usage Effectiveness (PUE) will be needed. Immersion cooling will likely become key to meeting these objectives.

Some liquid coolant manufacturers, like Perstorp, are incorporating synthetic esters. This molecular-level design approach allows for a balance of low viscosity, high flash point, high fire point, and high breakdown voltage, while being more environmentally friendly.

Immersion cooling fully submerges hardware in dielectric fluid for heat dissipation. Signal integrity, especially for high-frequency signals, is a critical factor that all immersion cooling system suppliers must carefully verify. AI servers have been submerged in cooling fluid for years, continuing to operate normally at the system level.

Intel’s SuperFluid technology, enhanced by Perstorp’s fluid, can significantly increase cooling capacity. Traditional single-phase immersion cooling systems typically manage around 500 watts per chip, but with this advanced solution, cooling capacity can reach up to 1,500 watts per chip, making it ideal for next-generation, high-end AI chips.

As the demand for more powerful AI servers, EV batteries, and EV charging cables continues to grow, the need for efficient and sustainable cooling solutions will become increasingly important. Liquid cooling technologies are poised to play a crucial role in meeting these challenges. Applications with higher power luminaires are also likely to increase.

More information is available here.

Image: Hon Hai Technology Group’s new liquid-cooled server racks, developed by subsidiary Ingrasys Technology Inc, are displayed at Nvidia Corp’s developers’ conference in San Jose, California, earlier this year. Photo courtesy of Hon Hai Technology Group via CAN.

David Shiller

David Shiller is the Publisher of LightNOW, and President of Lighting Solution Development, a North American consulting firm providing business development services to advanced lighting manufacturers. The ALA awarded David the Pillar of the Industry Award. David has co-chaired ALA’s Engineering Committee since 2010. David established MaxLite’s OEM component sales into a multi-million dollar division. He invented GU24 lamps while leading ENERGY STAR lighting programs for the US EPA. David has been published in leading lighting publications, including LD+A, enLIGHTenment Magazine, LEDs Magazine, and more.

See Full Bio