This NVIDIA Partner Promises 50% Reduction in Energy Consumption 

As the artificial intelligence boom continues to strain power grids, major tech companies like Microsoft, Alphabet, and Meta are investing heavily in data centre infrastructure to support their generative AI ambitions. 

However, this rapid expansion has led to a significant surge in energy consumption, with AI expected to represent about 19% of data centre power demand by 2028, prompting the need for more efficient cooling solutions. 

Sustainable Metal Cloud (SMC) is a company that has developed a groundbreaking immersion cooling technology that not only costs 28% less to install compared to other liquid cooling methods but also has the potential to slash energy usage by up to 50%. SMC’s co-founder, Rosenfield, emphasised that their technology enables high-density hosting for GPUs, a crucial requirement for platforms such as Nvidia’s Grace Blackwell.

SMC’s HyperCubes: Sustainable AI Factories

SMC operates cutting-edge ‘sustainable AI factories’ in Singapore and Australia, utilizing their proprietary HyperCubes. These innovative containers house servers equipped with Nvidia processors, which are submerged in a synthetic oil called Polyalphaolefin. 

This oil is remarkably effective at dissipating heat compared to traditional air cooling methods, allowing SMCs to significantly reduce energy consumption in their data centres.

The containerised nature of SMC’s solution offers the advantage of swift deployment in various locations, enabling the company to quickly establish new availability zones to meet client demand.

SMC is also expanding its reach into markets such as Thailand and India.

Partnerships and Investments

SMC has forged strategic partnerships with industry giants like Nvidia and Deloitte. As Nvidia’s preferred cloud partner for computing and AI, SMC provides GPU clusters designed by the chip manufacturer. 

In July, SMC announced a collaboration with Deloitte, granting the consultancy’s clients access to Nvidia’s GPU computing infrastructure for AI application development. The company has also attracted investment from Keppel ST Telemedia Global Data Centres, one of Asia’s largest data centre operators. 

Currently, SMC is in the process of raising $400 million in equity and $550 million in debt to expand its data centre operations beyond Singapore.

The Growing Trend of Liquid Cooling

As technology companies grapple with increasing performance demands and the risk of overheating, liquid cooling solutions are gaining traction in the data centre industry. Giordano Albertazzi, President of digital infrastructure provider Vertiv, predicted that the adoption of liquid cooling will accelerate in 2024. 

During Computex in June, Nvidia CEO Jensen Huang highlighted the benefits of Supermicro’s direct liquid cooling technology, claiming that it reduces energy consumption and lowers costs in AI data centres. Similarly, Supermicro CEO Charles Liang emphasized the superior power efficiency of liquid cooling, resulting in improved performance, reduced pollution, and lower energy expenses.

Challenges and Future Outlook

Despite the growing interest in liquid cooling, implementing these solutions comes with its own set of challenges. SMC’s Rosenfield pointed out that most data centres are not currently equipped to handle liquid cooling methods, whether immersion or direct chip cooling. 

The industry is still exploring the most effective ways to integrate these solutions, and it is likely that multiple approaches will emerge, such as hardware innovations, directed flow technologies, novel cooling fluids, and new approaches to system design. 

Vertiv’s Albertazzi noted that while air cooling remains prevalent in data centres and will continue to be used even in high-density AI environments, the shift towards liquid cooling is inevitable as the demand for more efficient and sustainable cooling solutions grows. 

Direct cooling can allow for highly efficient temperature management, reduce reliance on other cooling infrastructures like fans, and has the potential for waste heat reuse. 

As countries like Singapore invest heavily in AI initiatives and promote ‘green’ data centres, companies like Sustainable Metal Cloud are well-positioned to lead the charge in revolutionizing the way we cool our data centres, paving the way for a more energy-efficient and sustainable future in the AI industry. 

Despite AI being expected to account for 18.0% of the carbon emissions in data centres by 2027, it also has the potential to drive sustainability efforts and reduce overall emissions.