AI boom stresses the power infrastructure;

Nvidia partner claims it can cut data center energy demand by 50%.

 

 

 

 

• To support generative AI, large tech giants like Microsoft, Alphabet, and Meta are spending billions on data center infrastructure. However, the growth of data centers has also resulted in an increase in the demand for energy.

According to Sustainable Metal Cloud, its immersion cooling method can save up to 50% on energy use and can be installed for 28% less than existing liquid-based alternatives.

• It permits GPUs to be hosted at great densities. For platforms like [Nvidia’s] Grace Blackwell, it makes hosting possible that we need to see,” said Tim Rosenfield, Sustainable Metal Cloud’s co-founder and co-CEO.

The surge in artificial intelligence is driving up the requirement for processors with greater power and for the energy required to cool data centers.

Data center operator Sustainable Metal Cloud, which runs its HyperCubes in Singapore and Australia as “sustainable AI factories,” sees this as an opportunity.

With Nvidia CPUs, the HyperCubes’ servers are submerged in polyalphaolefin, a synthetic oil that dissipates heat more efficiently than air. The company claims that their platform may reduce energy consumption by up to 50% when compared to the traditional air cooling method typically used in data centers.

It makes high GPU density possible to host. Regarding Nvidia’s most current generation of AI graphics processors, Tim Rosenfield, co-founder and co-CEO of Sustainable Metal Cloud, said: “It enables the sort of hosting that we need to see for platforms like [Nvidia’s] Grace Blackwell.”

The Singapore-based business further stated that compared to rival liquid-based solutions, the installation cost of its immersion cooling approach is 28% lower. The HyperCubes are designed to fit into any kind of data center and can be deployed in empty places in already-existing data centers.

Because our system is containerized, we can move fast to any location. Additionally, we can create additional availability zones in response to client requests,” Rosenfield stated.

According to him, SMC is branching out into areas like Thailand and India.

Nvidia and Deloitte are already two of the company’s principal enterprise partners. GPU clusters created by the chip giant Nvidia are available through SMC, which is the company’s preferred cloud partner for computing and AI. SMC and Deloitte announced a cooperation in July whereby SMC will give its clients access to Nvidia‘s GPU processing infrastructure so they may develop artificial intelligence applications.

Data center demand has increased as a result of governments and corporations racing to capitalize on AI’s revolutionary potential.

In order to promote its aspirations for AI, countries like Singapore, where SMC is located, are also attempting to reduce the high energy usage by advocating for “green” data centers, to which the nation has contributed more than 500 million Singapore dollars ($379.7 million).

Temasek, a state investor in Singapore, has provided investment to Sustainable Metal Cloud. ST Telemedia Global Data Centres is one of the biggest data center operators in Asia.

According to Bloomberg, which cited people familiar with the subject, SMC is presently seeking $400 million in stock and $550 million in debt, with the proceeds going toward expanding its data center operations outside of Singapore.

Vigor of liquid cooling increasing

As overheating concerns rise along with increased performance requirements, technology organizations are increasingly considering liquid cooling for their data centers.

Large aisles are customary in data centers to facilitate the flow of chilled air because equipment failure and downtime might result from overheating. Liquid cooling should theoretically allow these facilities to accommodate more servers within their current footprint.

The CEO of Vertiv, a developer of digital infrastructure, Giordano Albertazzi, stated to CNBC in June that 2024 may see a surge in the use of liquid cooling. Vertiv offers entirely liquid-cooled data centers in addition to hybrid air- and liquid-cooled options for thermal control.

Nvidia CEO Jensen Huang promoted the advantages of Supermicro’s direct liquid-cooling technology at Computex Taipei in June, claiming that it reduces expenses and energy consumption in AI data centers.

According to Supermicro CEO Charles Liang’s said, liquid cooling provides greater power efficiency, which boosts performance while lowering energy costs and emissions.

Although there is excitement about liquid cooling, Rosenfield of SMC says there are still challenges.

Most data centers are not designed to manage two types of liquid cooling: immersion and direct chip cooling. Rosenfield predicts that the market will probably use a variety of techniques to figure out how best to use this.

“There is still a lot of air cooling that happens in the data center and will continue to happen even in the full high-density AI data center,” said Albertazzi of Vertiv.