Singtel has partnered with Nvidia to launch an artificial intelligence (AI) Center of Excellence (CoE) to help organizations overcome the infrastructure and skills bottlenecks that prevent them from scaling up AI deployments.
The multi-million dollar facility announced today is expected to provide an implementation pathway for organizations struggling to move their AI initiatives from the experimental stage to full-scale operations.
Bill Chan, CEO of Singtel Digital Infraco, the telco’s digital infrastructure business, said that unlike other AI CoEs in Singapore, Singtel’s CoE focuses on applied AI, with companies bringing real-world problem statements and working with an ecosystem of large-scale language model (LLM) manufacturers, application providers and system integrators.
To help customers scale up their AI deployments, Singtel has also designed a CoE to serve as a testbed that mirrors commercial infrastructure. Chan likened this to a national power grid consisting of AI data centers that act as generators, fixed networks that act as transmission lines, and edge locations that act as substations.
“Think of this center of excellence for applied AI as a micro-AI grid,” says Chan. “Not only do we experiment and resolve bottlenecks, but when we move into full-scale deployment, we seamlessly switch back to the main AI grid to obtain resources.”
Mark Hamilton, Nvidia’s senior vice president of solution architecture and engineering, explained that the partnership provides the five-layer foundation for AI deployment first advocated by Nvidia CEO Jensen Huang.
This foundation will consist of physical land, power and data center facilities provided by Singtel’s Nxera data center division, followed by Nvidia’s graphics processing units (GPUs). The third layer is the broader AI infrastructure, including networking and cloud orchestration, followed by AI models and finally applications.
Hamilton said NVIDIA plans to leverage its network of 40,000 AI startups to help Singtel customers build AI applications. He also highlighted the importance of open models such as Nvidia’s Nemotron for sovereign AI, ensuring that “Singapore’s data and unique competitive advantages remain in Singapore and are managed by Singapore companies.”
Additionally, Nvidia will work with Singtel to prepare data centers capable of handling extreme power densities without violating rigorous new sustainability metrics.
Singtel’s Nxera data center currently operates Nvidia GB200 Blackwell systems running at approximately 200 kW per rack. However, Chang noted that the CoE is already preparing for the 2027-2029 rollout of Nvidia’s next-generation Rubin Ultra chips.
“The next generation I’m talking about is 600 kW to 1 megawatt per rack. That’s insane in terms of power density, 60 to 100 times more than today’s average data center,” Chan said.
Hamilton pointed out that this extreme density requires less data center real estate. Blackwell chips are 50 times more energy efficient when running AI models than the previous generation Hopper, significantly reducing the physical space required for large-scale computing power. “Today’s GPU data centers look much more like basketball fields than most football fields,” Hamilton said.
The efficiency of AI data centers is critical as the Singapore government prepares to tighten regulations in this area. Later this year, the Singapore government is expected to table a new digital infrastructure law in parliament.
The bill seeks to establish baseline energy efficiency requirements for all data centers, including new and existing facilities, as well as mandatory cybersecurity measures and incident reporting requirements for major cloud service providers and data centers to ensure economic resiliency.
In addition to helping enterprises deploy and scale AI applications in AI data centers, the CoE will also focus on edge AI and low-latency 5G networks as the industry moves beyond generative AI to embodied, physical AI such as autonomous robotics, humanoids, and drone swarms.
The CoE will be located in the Punggol Digital Precinct. The district is currently being transformed into a school district-scale testbed for companies to trial and commercialize real-world robotics applications in real-world production environments.
Chang said the CoE is expected to open in about three months. In the meantime, large customers in sectors such as government, healthcare, banking, and transportation are already defining their use cases and leveraging Singtel’s existing GPUs to begin their AI journey.
