Current and Projected AI-specific Computing Power Available and Plans for National Cloud
Ministry of Digital Development and InformationSpeakers
Summary
This question concerns Singapore’s AI-specific computing power and plans for a national cloud to support research and education. Ms He Ting Ru inquired about current capacity, legal frameworks to mandate private data centre contributions, and comparisons with other leading AI nations. Minister for Digital Development and Information Mrs Josephine Teo highlighted Singapore's 1.4-gigawatt capacity and plans for 300 megawatts more, noting that AI compute fluctuates based on demand. She explained that research needs are met via commercial clouds and the National Supercomputing Centre’s ASPIRE 2A and 2A+ supercomputers, which provide 30 PFLOPS of power. The government prefers partnering with industry over mandating private contributions to remain responsive and support the National AI Strategy 2.0.
Transcript
22 Ms He Ting Ru asked the Minister for Digital Development and Information (a) what is the estimated total of AI-specific computing power currently available in Singapore's data centres; (b) whether there are plans to establish a national cloud for AI research and education; (c) if so, what percentage of this computing power does it aim to secure; (d) what legal frameworks are being considered to mandate contributions from private AI data centres to such a national cloud; and (e) how do these potential frameworks compare to those in other leading AI nations.
Mrs Josephine Teo: Singapore is a regional data centre hub with a total capacity exceeding 1.4 gigawatts. We have one of the highest concentrations of data centres in the region and our operational data centre capacity per capita exceeds that of regional markets, such as Beijing, Hong Kong, Seoul, Sydney and Tokyo. To support continued growth and innovation in Singapore, we aim to provide at least 300 megawatts of additional capacity in the near term and more through working with the industry to explore green energy deployments. The amount of AI-specific compute resources available in these data centres depends on the workloads that the private and public sectors expect to process and how these workloads are distributed. It changes dynamically in response to needs and available alternatives.
Our compute needs for AI research are being met through a combination of on-premise and commercial cloud capacities. This approach is more responsive to demand than mandating contributions from private data centres.
The National Supercomputing Centre (NSCC) is a significant contributor to the compute resources available for our research ecosystem, including for AI research. The launch of the ASPIRE 2A and 2A+ research supercomputers in October 2024, which have 30 PFLOPS of aggregated compute power, will help to address the growing demand for high performance computing resources, complement existing infrastructure and enable new research opportunities. NSCC is also developing the next supercomputer that will enhance Singapore's high performance computing capabilities to support national research initiatives.
Developing Singapore's compute infrastructure is vital to support our National AI Strategy 2.0 ambitions. We are actively working with industry partners to avail access to the compute resources needed to maintain our global competitiveness.