Singapore Built a $470M Supercomputer. Nobody Knows How to Drive It.

Technology
June 11, 2026

By now you have seen the CNA video. The sweeping shots of blinking server racks. The minister at the podium. The earnest researcher talking about cancer treatments. The ominous music about the “compute divide.”

It is a good video. It tells a great story.

It is just not the real one.

Let us talk about what Singapore actually bought with its half a billion dollars, what it can actually do, and why the most honest person in the entire saga might be the guy who runs the whole thing.

Part I: The Numbers They Told You

Here is what the brochure says.

• 180,000 plus cores. Your laptop has eight.

• 100 quadrillion calculations per second. That is 17 zeros.

• Four times the power of its predecessor.

• More than 1,500 projects across simulation, data analytics, and AI.

And here is what the brochure did not say.

That predecessor was Aspire 2A launched less than two years ago for 40 million dollars. They upgraded from a perfectly good system to a slightly more perfectly good system before anyone had really figured out how to use the first one.

That is like buying a Ferrari, leaving it in the garage because you have not taken driving lessons, and then buying a second Ferrari because the first paint job is technically last season.

Part II: The Numbers They Did Not Tell You

Let us pull back the curtain on the actual funding pipeline.

In 2019, the National Research Foundation invested 200 million dollars into NSCC. First tranche. Build things.

In 2020, the ASPIRE1 system got upgraded. Singapore also secured access to Japan Fugaku supercomputer then the world fastest because domestic capacity was already stretched thin.

In October 2024, the NRF allocated another 270 million dollars over three years. Second tranche. The official line: nurture specialists and build next generation supercomputers for AI workloads.

That brings the confirmed total to 470 million dollars in NRF grants alone, not counting whatever the individual systems actually cost on top.

Now here is the quote that should make you wince.

It comes from the NSCC’s Chief Executive, Dr Terence Hung, in October 2024 after the first 200 million had already been spent.

“There are fewer than 100 professionals proficient in HPC in Singapore.”

“It can take upwards of a year for programmers to come to grips with supercomputers.”

“We have users who can run applications, but we need users who can run it effectively.”

Part III: The F1 Analogy They Love

The NSCC CE used an analogy in the CNA video that was supposed to be reassuring. He said that giving an untrained researcher a supercomputer is like giving a normal driver an F1 race car.

This was meant to convey: do not worry, we are teaching people to drive.

What it actually conveys is: we built an F1 car before we had anyone who could drive one, and we are only now getting around to offering lessons.

Extend the analogy properly and it gets worse.

• The car is Aspire 2B. 180,000 horsepower. It exists.

• The drivers are likely fewer than 100 scarce specialists, most of whom already have jobs.

• The racetrack is PSA, Micron, DSTA organisations that quietly have their own cars and do not need yours.

• The spectators are the World Governments Summit, applauding from the stands.

• The bill is 470 million dollars, payable by the taxpayer.

The F1 analogy is not wrong. It is just incomplete. What it leaves out is this: what if the people who actually need to go fast already have their own cars?

Part IV: The Real Applications

The video talks about weather forecasting and cancer treatment. Let us be honest about what this machine will actually run.

Port and logistics simulation

Sounds boring but is genuinely valuable. PSA might use it but PSA already has their own clusters for operational work.

Defence and maritime security

Exactly the kind of sovereign workload that justifies domestic compute. It is also classified and will run on air gapped systems, not NSCC.

Sovereign AI training

The one clear win. Training models on Singapore specific data without sending it overseas is a genuine, uncontested need. GovTech, HTX, and CSA are the natural customers.

AStar academic research

The actual user base. Materials science, genomics, computational biology. Valuable work, but not the kind that generates direct GDP impact.

The pattern is obvious. The high value commercial use cases mostly do not need NSCC. They solved their compute problem their own way a long time ago. What NSCC gets is the research community. That is valuable, but it is not justify a half billion dollar cheque valuable.

Part V: The 15 Percent Utilisation Problem

Here is the inconvenient math.

A 180,000 core system running at 15 percent utilisation means the effective cost per compute hour is six to seven times higher than the sticker price. The fixed costs facility, cooling, staff, software licences do not shrink when nobody shows up.

The dynamics create a lovely doom loop.

1. Few people use the machine, so the per user cost is high.

2. High per user cost means NSCC has to charge high rates.

3. High rates mean researchers compare with alternatives and walk away.

4. Even fewer people use the machine. Repeat.

The NSCC CE acknowledged this problem indirectly when he said fewer than 100 people in Singapore can use HPC effectively. You do not need a crystal ball to see where this lands.

Part VI: The Sovereignty Argument

To be fair, there is a real argument for Aspire 2B. It is just not the one in the video.

The World Governments Summit report from May 2026 which I verified independently it exists, it is real, the CNA video did not fabricate it makes a genuine case.

It says that high performance computing is becoming as critical to future prosperity as food security and energy independence. It warns that nations which fall behind will find it increasingly difficult to compete economically or maintain the sovereignty of their physical and digital borders.

The report frames a compute divide as an existential threat. It flags supply chain vulnerability. It warns about dependency on foreign technology providers.

This is insurance. Strategic redundancy. A hedge against a future where the US China tech war escalates and suddenly Singapore cannot access foreign compute for sensitive workloads.

The problem is that you cannot sell we spent half a billion on insurance in a CNA news segment. You can sell we are curing cancer and predicting typhoons. One of these is honest. The other makes for good television.

The Burning Question

At 15 percent utilisation, and with fewer than 100 qualified users in the country after 470 million dollars of investment, will NSCC next funding tranche go to more hardware or will someone finally ask whether the problem was never the machine?