The global platform for building Artificial Intelligence (AI) that delivers high performance with low energy consumption, the Sustainable Metal Cloud (SMC) by Firmus, is now operational. Through innovative "immersion cooling" technology inherent in Firmus’ proprietary ‘Sustainable AI Factories', SMC halves the energy typically consumed in AI operation, significantly reducing CO2 emissions.
Beyond environmental benefits, this resource-smart solution also trims the cost of running AI workloads by as much as 70% compared to conventional cloud platforms. SMC provides easy access to substantial AI infrastructural capacity and can operate as a public or private cloud solution. It utilises NVIDIA H100 clusters for generative AI training on a large scale, in addition to CPU & GPU services for AI inference. In the face of growing concerns over the environmental impact of digital industries, the technology SMC brings forth guarantees a much smaller energy footprint than most existing cloud solutions.
SMC is already fully functional in Singapore with 1,200 NVIDIA H100 GPUs and powerful computer chips specifically intended for AI. In a bid to further extend its reach and impact, SMC is in the throes of swift expansion in collaboration with ST Telemedia Global Data Centres. Their ambitious projection is to have 5,000 GPUs operational in Singapore by the end of the year, as well as to roll out additional expansions in India, Thailand, Australia and Europe.
The advent and swift growth of AI hold the promise of immense benefits. Yet, these come with an important environmental drawback: the considerable energy demand from the computing power required to run AI applications. The significant issue stems from data centres, which are forecasted to consume around 4.5% of global electricity by 2030, according to the International Energy Agency (IEA). Similarly, current trends suggest that the insatiable demand for AI could push global electricity use up by an added 0.5% by 2027.
In Singapore, where SMC first launched its services, data centres are already responsible for over 7% of the national energy grid’s consumption. Because of the naturally hot and humid climate, the existing data centres operate relatively inefficiently. This presents an urgent need for AI solutions that abide by sustainability principles. SMC addresses this pressing issue by providing an AI platform that greatly reduces the environmental repercussions of AI development and operation.
Across two locations in Singapore, Loyang and Ayer Rajah, the Sustainable AI Factories, integral to SMC’s AI services, operate with almost half the energy usually required by run-of-the-mill cloud installations in the city-state. This energy-saving initiative is contributing significantly to CO2 emission reduction, not to mention the monetary savings. It also aligns well with the goals of the National AI Strategy 2.0.
Apart from sustainability and scale, another focus of SMC is cost-effectiveness, ensuring no compromise on performance. Leaders of conversation around energy efficiency in the AI sector, SMC’s team of researchers and engineers prioritise 'watts per token'. Through their Sustainable AI Factories utilising immersion cooling, SMC can cut energy usage for hosting and delivering AI GPU clusters by up to 50%, subsequently reducing CO2 emissions.
This commitment positions SMC as a key player within sustainable AI infrastructure. To enhance its offerings, the firm has formed partnerships with NVIDIA, Dell Technologies, and WEKA and also holds official cloud partner status with NVIDIA. Headquartered in Singapore, SMC is comprised of a growing team of AI experts and engineers and operates as part of a joint venture with ST Telemedia Global Data Centres.