Home**HPC**Liquid Cooling Means More Performance And Less Heat For Supercomputing

If there was ever a demonstration of Jevons’ paradox, it’s the supercomputing sector. According to this law of economics, consumption rises, rather than falls, with production efficiency. William Stanley Jevons conceived it in the 1860s as developments like the Watt steam engine made coal more efficient. Attracted to cheaper energy, industry used more coal, not less.

Similarly, no matter how much more efficient chips get, AI algorithms can never get enough compute density. But there’s a difference between coal engines and AI supercomputers: whereas heat was coal’s main output, it’s an unwelcome side effect of high-density compute. We’re reaching the point where conventional air conditioning can’t easily cope with it.

The heat question is weighing heavily on companies even as they pursue lofty AI goals. The calculus has shifted. According to the IDC and Lenovo CIO Playbook 2025, sustainability now ranks among the top business priorities for infrastructure buyers.

That’s where Lenovo’s Neptune direct water cooling system comes in. It pipes water straight to the server nodes themselves, pulling heat directly off CPUs, GPUs, and memory modules before it ever reaches the surrounding air.

Neptune operates at roughly 45° C, far warmer than the 18° C chilled water loops most facilities rely on. That temperature difference eliminates traditional datacenter chillers entirely. The sixth-generation systems can now remove up to 100 percent of server heat through the liquid loop alone, leaving fans with almost nothing to do.

Lenovo Takes Home Both Top500 And Green500 Awards

Anyone who follows supercomputing knows the Top500 – a bi-annual ranking of the world’s fastest machines, first published in June 1993 at the Mannheimer Supercomputer Seminar in Germany. Jack Dongarra of the University of Tennessee developed the LINPACK benchmark that still underpins it today.

The Green500 came later, launched in 2007 by Kirk Cameron and Wu-chun Feng at Virginia Tech to measure something the original list ignored: energy efficiency, expressed in gigaflops per watt. The two merged in 2016, meaning a single submission now covers both rankings.

Between them, these lists measure raw speed and efficiency, which are two metrics often in contention. Dominating both is exceptionally rare. As of November 2025, Lenovo holds the number-one position on each.

The ThinkSystem SR780a, running Neptune cooling, has reached a Power Usage Effectiveness of 1.1 (1.0 would be theoretically perfect). That means almost nothing is lost to cooling overhead.

This isn’t Neptune’s first official recognition among green tech experts. It also picked up CRN’s Best Green Energy Product award for 2024. And multiple Lenovo platforms such as the SD650, SR670, and SD665, have landed in the Green500’s top ten.

Liquid Cooling In Action

Companies with big compute density demands are taking notice. DreamWorks Animation saw a 20 percent performance jump on Neptune-powered clusters while actually cutting cooling requirements, while meteorological agencies in Korea and Malaysia rely on the same technology to run weather forecasting models.

Neptune systems draw about 40 percent less power than their air-cooled counterpart. That feeds directly into Lenovo’s broader pledge – net-zero carbon emissions by 2050, validated by the Science Based Targets initiative.

For buyers, lower energy bills and simpler cooling infrastructure mean a noticeably lighter total cost of ownership. So while organizations might follow the Jevons principle by piling on compute density, Neptune saves them more money with every new chip it cools.

Sponsored by Lenovo.

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between. Subscribe now