What are GPUs?
A Graphics Processing Unit (GPU) is a specialised electronic circuit that can perform mathematical calculations at high speeds to handle applications like machine learning.
What is PUE?
Power usage effectiveness (PUE) is a ratio that describes how efficiently a computer data centre uses energy; specifically, how much energy is used by the computing equipment. The closer the PUE is to 1.0, the more efficient the DC is.
As governments and businesses harness the transformative impact of Artificial intelligence (AI), they are also looking to mitigate the hefty energy requirements of powering their AI ambitions.
Singapore is doing the same. Alongside the country’s National AI Strategy 2.0, Singapore's 2024 Budget outlined plans to invest up to S$500 million in high-performance compute resources. And with the growing use of AI and autonomous and robotic systems propelling the demand for compute resources, Singapore is also creating a sustainable pathway for the continued growth of Data Centres (DCs) in Singapore.
In May, Singapore released its new Green Data Centre Roadmap that aims to provide at least 300 megawatts of additional capacity in the near term—adding to the current 1.4 gigawatts of computing capacity in more than 70 cloud, enterprise, and co-location DCs on the island—with much more capacity unlocked through green energy deployments.
There are also plans to install best-in-class technologies to maximise the energy efficiency of all DCs in Singapore.
DCs are indeed the backbone of the AI revolution. Yet, their energy demands are significant, and this trend is expected to accelerate with the introduction of the next generation of Graphic Processing Units (GPUs), like NVIDIA’s Blackwell series – requiring up to 40 per cent more power than prior-generation Hopper GPUs.
An effort to enhance sustainable DCs is already taking place in Singapore with Sustainable Metal Cloud (SMC), the Singapore-founded and headquartered AI GPU cloud service, that utilises advanced and highly energy efficient GPU data centre hosting technology developed by its parent company, Firmus Technologies.
What is an LLM?
A Large Language Model is an artificial intelligence (AI) program that can generate and recognise text. These LLMs, such as programming languages, have been trained on large amounts of data and are built through machine learning. Inferencing is the process where live data — in the most common end-user case is a text-based query — is put through a trained AI model to make predictions or solve a task.
What are Sustainable AI Factories?
AI factories are facilities designed to meet the specific requirements of AI, providing infrastructure and resources to carry out complex AI-powered applications or models. Through its Sustainable AI Factories, SMC, an official NVIDIA Cloud Partner, uses advanced technology to overcome the challenge of power scarcity and infrastructure scalability for next-generation GPUs and provides universal access to sustainable AI solutions.
Using the HyperCube — a form of advanced immersion cooling technology where servers are submerged— SMC has achieved a Power Usage Effectiveness (PUE) of 1.031. Combined with further efficiencies at the computer level, GPU services delivered by SMC – in Singapore – use up to 50 per cent less energy per hour1 to develop and host AI workloads.
With these breakthroughs, SMC is able to provide organisations in Singapore with access to high-performance, cost-effective AI infrastructure at scale and do so from land and water-constrained Singapore.
Developing, fine-tuning, or inferencing Large Language Models (LLMs) that power the AI economy requires access to increasingly large quantities of high-performance GPUs.
As the demand for the AI economy grows, securing these specialist computers and making them accessible to companies and countries is becoming a barrier to innovation. This goal is not easily obtained as – in many cases – there is a substantial lack of available power or suitable infrastructure that these powerful computers require.
With EDB’s support, SMC has prioritised deployments of its GPU cloud clusters in Singapore – with 1,600 H100 SXM GPUs today, and up to 4,000 by 2024 – to accelerate Singapore’s AI capabilities.
SMC established its global headquarters in Singapore based on several strategic advantages the country offers. Singapore is a growing global DC hub with a vibrant ecosystem offering infrastructural, geographical, talent, and technological advantages.
“As we continue to expand our AI cloud in Singapore and beyond, SMC remains dedicated to powering the AI revolution sustainably,” said Tim Rosenfield, Co-CEO of SMC.
"Our vision and commitment align perfectly with Singapore’s forward-thinking policies, such as the National AI Strategy 2.0 and Singapore Green Plan 2030; representing a win-win opportunity for both parties to redefine the global landscape of AI and sustainability."
SMC is also collaborating closely with EDB to launch the GreenFutures Accelerator, an initiative that provides a sustainable AI platform that propels local startups and scale-ups seeking to harness the power of GenAI cost-effectively and sustainably.
Startups can get access to:
Find out how you can be a part of this program to realise your sustainable AI ambitions:
1 Sustainable Metal Cloud. An entirely sustainable AI infrastructure stack. Accessed at: https://smc.co/sustainability
