Singtel and NVIDIA want to bridge Singapore's AI pilot-to-production gap

A new Centre of Excellence for Applied AI aims to address a persistent problem: enterprises that can experiment with AI but struggle to deploy it at scale.

Singapore has no shortage of organizations dabbling in AI. What it has a shortage of is organizations that can move past the dabbling. That's the problem Singtel's Digital InfraCo is trying to solve with a newly launched Centre of Excellence (CoE) for Applied AI, built in partnership with NVIDIA.

Announced on 24 February, the CoE is pitched as a structured environment where enterprises and government agencies can bring real-world problems, test solutions with actual infrastructure, and–critically–develop a credible path to full-scale deployment rather than another pilot that quietly disappears.

Bill Chang, CEO of Singtel's Digital InfraCo, put it plainly, "We have seen many organizations struggle to move from the excitement of AI to meaningful impact." It's a candid acknowledgement that the market has an adoption problem, not just an infrastructure one.

The CoE will combine NVIDIA's platforms, models and technical specialists with Singtel's RE:AI sovereign AI cloud, Nxera data centers and network infrastructure. The idea is to give organizations hands-on access to the full stack–compute, software, expertise–without having to procure or configure any of it themselves first.

For channel partners and solution providers serving enterprise clients in Singapore, the significance is less about the headline partnership and more about what the CoE's focus areas reveal about where demand is actually heading.

The four pillars are telling: next-generation data center design accommodating over 600KW power density per rack for large language models; an ecosystem of AI models, applications and solution providers; Edge AI capabilities tied to low-latency networks for use cases like robotics and autonomous systems; and applied AI talent development embedded into R&D projects.

The 600KW per rack figure is worth pausing on. Most enterprise data centers today operate at a fraction of that density. Building for this level signals that Singtel is positioning Nxera's facilities not for today's AI workloads, but for the compute-intensive models that are already moving down the pipeline.

It also reflects a broader industry reckoning with the infrastructure gap between what next-generation AI demands and what most facilities were built to handle.

NVIDIA's Ronnie Vasishta, Senior Vice President of Telecommunications, framed the collaboration around the progression from trials to scaled deployments–a sequence that sounds straightforward but has tripped up more organizations than most would admit.

The CoE also has a policy backdrop worth noting. Singapore's Budget 2026 flagged AI as a strategic national priority, with specific AI Missions targeting key sectors. The timing of this launch isn't coincidental, and Singtel's positioning around sovereign AI–keeping data onshore, governed under local requirements–is a deliberate play for public sector contracts that will almost certainly follow that budget commitment.

Whether the CoE translates into meaningful production deployments, or becomes another well-resourced initiative where enterprises come to learn and leave without changing their operations, remains to be seen. But the structure–real infrastructure, real NVIDIA specialists, real problem statements–at least gives it better odds than most.