Unisys has joined forces with Refroid Technologies to help data center operators install and maintain liquid cooling equipment.

The partnership places particular emphasis on facilities optimized for AI workloads.

Unisys traces its roots to a group of companies that made typewriters, calculators and other analog hardware in the nineteenth century. The firms became a single organization through a series of mergers that unfolded over a century. Unisys took on its current brand in 1986 and moved into technology consulting about a decade later. Today, that market accounts for practically all its revenue.

Refroid is an India-based manufacturer of liquid cooling equipment. According to the companies, their partnership will combine the hardware with professional services from Unisys. The consulting firm will help clients install Refroid’s gear in their data centers. Additionally, Unisys can take over ongoing equipment maintenance tasks. It will use AI to support the work of its technicians.

“When people read ‘AI-powered service management,’ they sometimes picture intelligence baked into the cooling hardware itself,” Vivek Swaminathan, the director of products and solutions for digital workplace services at Unisys, told Boardroom Insight. “The hardware is Refroid’s engineering. The AI is in how Unisys runs the service operation around it. The hands-on work — installing, commissioning, validating these systems — is done by Unisys global field engineers trained and certified on Refroid’s equipment; that’s the workforce enablement piece.”

One of Refroid’s specialties is making cold plate cooling equipment. The basic idea behind the technology is to integrate metal plates with tiny embedded pipes into data center racks. The embedded pipes circulate cooling liquid, usually water, that absorbs heat from server components. The water is then channeled outside the servers by pumps.

Refroid also makes immersion cooling gear. The technology works by submerging a server’s key components in a non-conductive coolant. That approach removes heat more efficiently than cold plates, but it can be significantly more expensive. As a result, adoption of the technology is currently fairly limited.

Unisys says that its AI automation software kicks in the moment a cooling system experiences a technical issue. “Unisys’ AI handles everything around that visit: it reads an incoming request, works out what’s actually needed, and gets the right certified engineer on-site with the right parts and knowledge, automatically, across regions and time zones,” Swaminathan explained. “The next step as deployments scale is tying into telemetry from Refroid’s remote monitoring tools, so Unisys can flag issues before they cause downtime instead of after. Cooling a GPU rack is a physics problem. Keeping that infrastructure healthy globally, for years, is a coordination problem — and that’s where AI earns its keep.”

Unisys and Refroid are going after a large and growing market. Gartner estimates that spending on AI data center infrastructure will exceed $1.4 billion this year.

Photo courtesy of Unsplash