This Nvidia Partner Could Help AI Data Centers Cut Power Consumption in Half

The AI ​​boom is driving demand for more powerful processors as well as the power needed to keep data centers cool.

This represents an opportunity for data center company Sustainable Metal Cloud, which operates “sustainable AI factories” made up of HyperCubes in Singapore and Australia.

HyperCubes contain servers with Nvidia processors immersed in a synthetic oil called polyalphaolefin that draws heat more efficiently from the air. The company says its platform reduces power consumption by up to 50 percent, compared to traditional air-cooled technology typically used in data centers.

“It enables high-density hosting of GPUs,” said Tim Rosenfeld, co-founder and co-CEO of Sustainable Metal Cloud, referring to the new generation of AI-powered GPUs. “It enables the kind of hosting that we need to see for platforms like Nvidia’s Grace Blackwell.” Nvidia It was announced in March.

The Singapore-based company also said its immersion cooling technology is 28% cheaper to install than other liquid-based solutions. HyperCubes are designed to fit into any data center and can be deployed in unused spaces within existing data centers.

“Our solution is container-based, which means we can get anywhere very quickly. We can also open new availability zones in response to customer demand…,” Rosenfeld said.

He added that SMC is expanding into other markets such as Thailand and India.

The company already counts Nvidia and Deloitte among its key partners. SMC is Nvidia’s preferred cloud partner for computing and AI, offering GPU clusters designed by the chip giant. In July, SMC announced a partnership with Deloitte that will provide access to Nvidia’s GPU computing infrastructure for consulting clients building AI applications.

Governments and businesses have rushed to capitalize on the transformative impact of AI, and as a result demand for data centers has soared.

Countries like Singapore, where SMC is headquartered, are also looking to mitigate their massive energy consumption by pushing for “green” data centres to support their AI ambitions, with the country committing more than S$500 million (US$379.7 million).

Sustainable Metal Cloud has secured funding from ST Telemedia Global Data Centres, which is backed by Singaporean state investor Temasek and is one of the largest data centre operators in Asia.

SMC is currently raising $400 million in equity and $550 million in debt, with the money going to expand its data center outside Singapore, Bloomberg reported, citing people familiar with the matter.

Technology companies are increasingly looking to liquid cooling for their data centers as increased performance needs also increase the risk of overheating.

Traditionally, data centers have large aisles to allow cooled air to circulate as overheating can cause equipment to fail and shut down. In theory, liquid cooling should enable these facilities to cram more servers into their existing space.

Giordano Albertazzi, CEO of digital infrastructure provider Vertiv, told CNBC in June that liquid cooling adoption could accelerate in 2024. Vertiv’s thermal management offerings include hybrid air-liquid cooling, as well as fully liquid-cooled data centers.

At Computex Taipei in June, Nvidia CEO Jensen Huang touted the benefits of Supermicro's direct liquid cooling technology, which he said saves power and reduces costs in AI data centers.

Liquid cooling is more energy efficient, resulting in better performance, less pollution and lower energy costs, Supermicro CEO Charles Liang told CNBC in June.

Despite the hype surrounding the deployment of liquid cooling technology, challenges remain, according to SMC's Rosenfeld.

“Most data centers are not ready to use any kind of liquid, whether it’s immersion or direct cooling of the chips,” Rosenfeld said. “The market is trying to figure out the best way to use this and I think there will be multiple ways.”

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