UNICOM Engineering has expanded its partnership with GRC (Green Revolution Cooling) into pre-integrated immersion-cooled compute that arrives as one order rather than as a server purchase and a cooling purchase the buyer has to make agree. The two companies say the arrangement covers AI, HPC, GPU as a Service, high-density compute and what they call Industrial Data Centers, deployed on factory production lines, remote depots and distributed warehouse environments.
Single-phase immersion submerges the server in a dielectric fluid instead of pushing air across it. GRC's release states that its technology eliminates chillers, computer room air conditioners and air handlers. That is the trade a buyer is being asked to weigh: the mechanical plant a comparable air-cooled hall would need does not get built, and a fluid, a tank and a fluid management system do.
The edge case is the more specific one. Both companies point to telecom sites, retail chains, healthcare and other distributed locations where a small communications closet cannot support AI-driven applications such as video analytics, and they name GRC's ICEraQ Nano as the unit for those rooms. GRC's own launch release for the Nano puts it at 10U, up to 13 kW of cooling, liquid-to-air with no chilled water required, shipped pre-filled with the company's ElectroSafe fluid and rated for a 15-year system lifecycle.
"Together with GRC, we’ve created a solutions approach that removes the friction customers typically face when evaluating or deploying immersion cooling," said Rusty Cone, General Manager of UNICOM Engineering. What the release does not say is what the integrated systems cost, when they become orderable, which server platforms are in the catalogue, or what heat load the joint configurations are rated to remove.
Behind the announcement sits four years of joint testing and validation. GRC's water-free ICEraQ Nano will now be installed in UNICOM Engineering's lab and manufacturing centre in Plano, Texas, which the release presents as part of simplifying customer evaluation. The two firms also say they will align co-marketing, co-selling and co-development across AI, GPUaaS, industrial automation, research, defence, energy, telecom and distributed edge work.
Peter Poulin, chief executive of GRC, put the buyer benefit this way in the release: "Customers gain a fully engineered, tested, and supported system with integrated cooling to help them accelerate AI initiatives with confidence." Warranty and support terms for the combined system are not stated, and the release does not say which of the two companies holds the contract when a fault falls between the compute platform and the tank.
No customer is named and no deployment figure is given for the joint offering. GRC states its systems have been deployed in 24 countries, a company figure covering its whole installed base rather than anything specific to this partnership.
For buyers, the change here is procurement rather than thermodynamics: one supplier, one validated configuration and one lifecycle service relationship in place of two orders that have to be reconciled on site. Price, availability and a rated capacity for the integrated configurations are all still missing, and until they are published no like-for-like comparison against a direct-to-chip loop can be made.
Leave a Reply