Singapore has turned Aspire 2B into a national bet on faster AI research, better climate modelling and healthcare tools built closer to home.
Singapore’s newest supercomputer is not just another piece of government infrastructure. Aspire 2B gives the country’s researchers a much larger local engine for training AI models, running climate simulations and testing ideas that previously had to be simplified, delayed or sent overseas.
Launched on June 8 at Nanyang Technological University’s Innovation Centre, the system is operated by the National Supercomputing Centre Singapore and is now the country’s most powerful national research supercomputer. According to The Straits Times, it has more computing power than 120,000 high-end AI laptops combined, which is the sort of comparison that makes the point clearly enough: this is a scale play.
The practical audience is large. More than 9,000 public researchers across universities, research institutes and government agencies can tap the system to build more complex models and shorten experimentation time. That matters because modern AI research has become deeply dependent on access to compute. Good ideas still matter, but without enough processing power, many of them stay stuck at prototype stage.
Aspire 2B is built with 1,536 Nvidia H200 GPUs, each designed for the kind of heavy AI and scientific workloads that now sit at the centre of national research strategies. NSCC says the system can deliver up to 115 petaFLOPS, or more than 100 quadrillion calculations per second.
That is four times the computing power of Aspire 2A and Aspire 2A+ combined, the two national systems launched in 2024. It is also a major jump from the original Aspire 1, which came online about a decade ago and helped establish Singapore’s shared high-performance computing base.
For researchers, the upgrade changes what can be attempted locally. Climate teams can combine AI methods with physics-based simulations at higher resolution. Healthcare researchers can train models on richer clinical datasets. AI teams working on language models can move beyond broad English-first systems and build tools that understand the way people actually speak across Singapore and Southeast Asia.
That last point is not a minor feature. A*STAR’s MERaLiON model, developed with earlier NSCC resources, is already being used to understand regional languages and local speech patterns, including Hokkien, Mandarin, Tamil and Malay. One use case highlighted in Singapore involves Lion Befrienders, where the model helps automate routine check-in calls with seniors. With Aspire 2B, researchers can train larger multimodal models that work with text, images and audio, and that can handle underserved languages with more confidence.
Climate and healthcare are the clearest tests
Singapore has obvious reasons to care about better weather and climate modelling. A dense island city cannot treat rising seas, extreme rainfall and coastal protection as abstract concerns. The National Environment Agency’s Third National Climate Change Study used NSCC computing resources for high-resolution modelling, and the next stage is to make those simulations sharper and more useful for planning.
This is where supercomputing becomes less remote than it sounds. Better models can help decide where coastal defences go, how drainage systems are designed and how urban development is planned. The benefit is not a faster spreadsheet. It is better judgement before expensive, long-term infrastructure decisions are locked in.
Healthcare may be just as important. The Singapore Medical Foundation AI Model, launched in 2025, is expected to use Aspire 2B to train larger models from more diverse health records and clinical datasets, including text, photos and audio. The goal is earlier risk detection and more personalised preventive care, which is exactly where AI could make a useful difference if hospitals can trust the models and the data behind them.
Industry also has a place in this story. Maritime engineering firm Mencast used NSCC computing resources to test more than 10,000 marine propeller design variations within days, rather than taking weeks or months to work through far fewer options. That is the commercial side of the same argument: when experimentation gets cheaper and faster, companies can try more serious ideas before committing capital.
The system also sits inside a broader national AI push. Singapore’s National Research Foundation committed S$270 million in 2024 to build national AI infrastructure, and Aspire 2B is part of that effort. The country has also been expanding AI Centres of Excellence and investing in models such as SEA-LION and MERaLiON that reflect regional needs rather than simply importing tools trained for other markets.
There is a quantum angle coming as well. Aspire 2B is expected to be linked later in 2026 with Helios, a quantum computer being set up in Singapore through a partnership involving Quantinuum and the National Quantum Office. The near-term impact should be kept in perspective, but hybrid quantum-classical computing could eventually help in areas such as molecular simulation and advanced materials research.
The real measure of Aspire 2B will not be the size of the machine. It will be whether Singapore turns access into output: better forecasts, stronger medical models, more useful regional AI systems and industry tools that leave the lab. Compute is now a strategic resource. Singapore has bought itself more of it, and the next test is how well its researchers use the advantage.
Also read: Washington puts Alibaba, BYD and Baidu deeper into its China tech fight • ASML is pushing back as Brussels tries to steer Europe's chip future • Anthropic makes AI chips Wall Street's newest collateral
