Two-phase cooling systems target AI accelerator thermal challenges

Startup claims 14°C temperature reduction using refrigerant-based cooling, addressing GPU thermal density in AI infrastructure.

Abstract flowing shapes representing cooling system technology
AI-generated illustration · Sylvaris

Thermal Density Problem

Accelsius, an infrastructure startup, has developed two-phase cooling technology for AI accelerators that converts liquid-cooled Dell PowerEdge servers to run on refrigerants instead of water-based systems. The company claims temperature reductions up to 14 degrees Celsius compared to conventional liquid cooling.

The approach addresses increasing thermal challenges as GPU power density rises in AI training and inference workloads. Modern AI accelerators generate significantly more heat per square inch than previous server generations, pushing the limits of traditional cooling methods.

Infrastructure Implications

Two-phase cooling uses refrigerants that evaporate and condense in a closed loop, carrying heat away from processors more efficiently than single-phase liquid systems. The technology has existed in high-performance computing contexts but faces new adoption pressure from AI workload density.

Data center operators running large GPU clusters report thermal management as a primary constraint on deployment scale. Improved cooling efficiency directly affects the number of accelerators that can be installed in existing facility footprints.

sources
more in Cloud & Infrastructure
Google Cloud becomes Alphabet's fastest-growing segment at over 20% of revenue Cloud infrastructure now represents more than one-fifth of Alphabet's total revenue and operating profit, marking a strategic shift beyond advertising dominance. US Army exhausts annual AI token allocation, implements usage limits Rapid AI adoption in military operations faces practical constraints when organizations underestimate infrastructure requirements. AMD and Anthropic reach $5 billion AI infrastructure deal AMD's largest AI infrastructure commitment to date signals intensifying competition in the data center GPU market dominated by NVIDIA.