Stack Infrastructure has said it is expanding its AI data center capabilities with high-density racks to help support the growing requirements of machine learning workloads.

The data center developer said it will support high-density workloads via a closed-loop water cooling system.

Stack Infrastructure
– Stack Infrastructure

Stack’s offering includes customizable solutions that can support up to 30kW per rack with traditional air cooling; up to 50kW per rack with rear door heat exchangers; and up to 100kW per rack with direct-to-chip liquid cooling. The company said it intends to support 300kW or higher per rack with immersion cooling in the future.

Three hundred kilowatts represents an incredibly high density level, with only a handful of other companies currently offering support on that scale.

In August 2023, CyrusOne launched a new AI-specific built-to-suit data center design, using immersion cooling, and other techniques, to achieve high power densities. That same month, Digital Realty launched a high-density colocation service in 28 markets to support workloads of up to 70 kilowatts per rack, while Equinix announced in December it would be rolling out support for liquid cooling across a large proportion of its data center footprint. However, the company did not confirm what density its offering would be able to support.

“Stack was built for the world’s largest innovators, and from the beginning, we have prioritized the long-term scalability of our clients with a flexible data center design,” said Matt VanderZanden, chief operating officer of Stack Americas.

“As an AI-Ready digital infrastructure company with a proven track record of high-density deployments, Stack is building upon this foundation to continually address evolving client needs,” he said.

In North America, the company has operations in Atlanta, Calgary, Chicago, Dallas-Fort Worth, New Albany, Northern Virginia, Phoenix, Portland, Silicon Valley, and Toronto. In EMEA, it has a presence in Copenhagen, Frankfurt, Geneva, Milan, Oslo, Stockholm, and Zurich. Over in Asia Pacific, it has data centers in Canberra, Melbourne, Osaka, Perth, Seoul, and Tokyo.

The company said that it planned to expand aggressively to meet growing artificial intelligence demand, raising more than $290m in debt financing in November 2023 to fund the building of a dozen new data centers across the globe.