Liquid Cooling for AI & HPC
Overcome the thermal challenges of modern AI and high-performance computing with advanced liquid cooling solutions that maximize performance, efficiency, and scalability.
AI is hotter than ever. With AI servers that consume over 14kW at peak load on the roadmap, enterprises, colocation providers and hyperscalers all need proven cooling technologies to support the challenging heat production caused by AI training and inference clusters that can exceed more than 100kW per rack. At JetCool, our microconvective direct liquid cooling (DLC) for AI, offered in three form factors, gives our customers a wide range of technologies tailored for specific demands: SmartPlate System, SmartPlate, and SmartLid.
Faster Processing Speed
Cooling Energy Savings
TDP Support
SmartPlate™ Cold Plates
SmartPlates, our fully-sealed cold plates, efficiently cool the industry’s highest power AI devices, clocking in at over 3,000W TDP, and popular in hyperscale deployments. By utilizing hot inlet coolant temperatures above 50°C, SmartPlates enable facilities to eliminate the need for chillers and cooling towers, all while maximizing compute performance, processor lifespan, and power efficiency. Our patented microconvective liquid cooling technology allows SmartPlates to outperform microchannels by up to 40%, delivering a 37% temperature reduction.
SmartPlate™ System
SmartPlate System is a self-contained, plug-and-play liquid cooling solution for Dell PowerEdge servers. Integrated into multiple server form factors, it meets diverse AI requirements efficiently. Achieving up to 50kW rack densities without requiring a CDU, SmartPlate System saves an average of 13% total server power. This makes it a perfect choice for distributed AI at the edge or in power-constrained environments.
SmartLid™ Cooling Modules
SmartLid is a favorite for semiconductor manufacturers, chipmakers, and foundries, including the leading designers of CPUs and GPUs. A liquid-to-chip solution that eliminates all thermal pastes and interface materials, it supports over 2,000W in a single socket. This revolutionary cooling approach boasts an impressive heat transfer coefficient 10X greater than other methods making it the perfect choice for devices striving to be more powerful, efficient, and compact. SmartLid deployments in federal labs are improving efficiencies with coolant temperatures over 60°C and eliminating water consumption.
Maximize Compute
With JetCool’s liquid cooling modules, you can add more HPC to your cluster, increasing rack density. Put more compute performance in a smaller space. It’s that simple.
Eliminate Hot Spots
Eliminate hot spots and thermal variations with more efficient cooling technology. Remove locally elevated temperatures across the chip to increase performance.
Warmer Coolants
Processors that use JetCool’s microconvective direct-to-chip cooling modules can run the same compute-heavy workloads with higher coolant temperatures, over 60°C.
Reduce TCO
Eliminate power- and water-intensive infrastructure like chillers and cooling towers reducing on-site energy consumption by up to 18% and enable zero water consumption