Supermicro is planning additional facilities to meet increasing demand for liquid-cooled datacenter infrastructure, and will focus on delivering entire plug-and-play installations to customers.

The San Jose-based server provider to the hyperscalers said it continues to expand in Silicon Valley with three new campuses due to the demand for liquid cooling inside datacenters. This is partly driven by the AI craze, which is leading to beefier systems being procured that sport multiple hot-running GPUs.

According to Supermicro, these campuses will form part of a new liquid-cooled ecosystem, intended to deliver kit faster to customers worldwide. The facilities will focus on delivering entire plug-and-play liquid-cooled solutions, ranging from systems to racks to water towers, it claims.

"Many datacenter owners are looking for electricity-saving Direct Liquid-Cooled (DLC) solutions," claimed president and CEO Charles Liang. "Supermicro is developing building block liquid-cooled solutions for AI factories and the HPC market."

Those building block solutions include racks pre-configured with kit optimized for enterprise, high performance compute (HPC), or AI/LLM scenarios, with liquid cooling ready integrated.

Liang repeated his assertion that liquid-cooled datacenters will grow from being less than 1 percent of the market to about 30 percent of all installations over the next two years.

"This expansion positions us to capture the majority share of that growth," he stated.

We asked Supermicro how much it was investing in the new facilities and when it expects them to come online. A spokesperson said: "Our global expansion is both in San Jose, California, and Asia.

"In San Jose, earlier this year, Supermicro acquired approximately 20 acres of land and buildings that will be used for manufacturing, assembly, and warehouse facilities. Additionally, Supermicro is leasing more multi-use factory/warehouse space at another San Jose location.

"Overseas, Supermicro is constructing a new state-of-the-art multi-building facility in Malaysia. It will be part of Supermicro's global manufacturing and assembly sites, which can offer customers the ecosystem of plug-and-play liquid-cooled solutions, from systems to racks to water towers, reducing time to delivery."

That assertion about the market for cooling was made during the Computex trade show in Taiwan earlier this month, when Liang claimed that Supermicro was gearing up to build 1,000 racks a month equipped with its DLC technology and would be able to dispatch orders just two to four weeks after receiving them.

These expectations were supported by a Register survey back in April, which found that more than a third of enterprises (38.3 percent) expect to employ some form of liquid-cooled infrastructure in their datacenters by 2026.

New datacenters will run more efficiently, reducing their carbon footprint compared to air-cooled facilities, with significant operational savings realized through lower electricity use over time, Liang claimed.

Supermicro said that it continues to work closely with operators to match the right server technology to their workloads, which may be unique to the organization or cloud service provider.

Many of its server models are designed for liquid cooling, such as the 8U-20-node SuperBlade, claimed to offer the highest CPU and GPU density available, the 2U-4 node Supermicro BigTwin and the 4U-8 node FatTwin multi-node servers.

According to analyst outfit Omdia, Supermicro grew its share of server market revenue from 5 to 10 percent over the course of 2023 based on AI server sales, overtaking HPE during Q4. ®

https://www.theregister.com//2024/06/19/supermicro_liquid_cooled_facilities/