In a fireside chat titled “CEO Insights: Rewiring Asia’s Digital Infrastructure”, Singtel Digital InfraCo CTO Manoj Prasanna Kumar joined moderator Carl Roberts, Partner at Hadaara Consulting, to unpack how AI, cloud, connectivity, and APIs are converging to redefine the region’s digital backbone.
Across a fast-moving session, Kumar laid out an ambitious yet pragmatic blueprint for scaling AI across Asia, one that sees Singtel’s focus on connectivity, digital services, and digital infrastructure serving sovereign, enterprise, and regional AI requirements in Asia.
Building AI from the ground up: Data centres, chips, connectivity
Kumar opened with a clear thesis: AI infrastructure is built on three inseparable pillars: data centres, chips, and connectivity, and Singtel is deliberately interweaving all three.
Singtel’s early move into liquid-cooled, high-density AI-ready data centres has already enabled sovereign-grade AI cloud services for government, enterprise, and research organisations. With AI chip shortages and global hyperscalers prioritising the US and Europe for new deployments, Singapore risked becoming an AI-capacity afterthought.
“We saw a gap in the market,” Kumar explained. “Hyperscalers weren’t going to bring high-density GPU capacity here fast enough. We stepped in early so Singapore could emerge as an AI-ready nation.”
From this foundation, Singtel built RE:AI, its sovereign AI cloud platform, across three layers:
- IaaS, offering access to GPU clusters where customers can bring their own models and code.
- PaaS provides developer tools, data science environments, and low-code capabilities.
- MaaS or Models-as-a-Service, where Singtel partners with global model makers to deliver vertical solutions, from agentic AI platforms to video analytics services.
This vertically integrated approach, Kumar noted, enables enterprises to adopt AI in weeks rather than months and ensures low-latency access via Singtel’s network and subsea cables.
Engineering for milliseconds: The role of Paragon and the intelligent network
A major theme of the discussion was latency, a critical enabler as Asia embraces robotics, drones, vision systems, and physical AI.
To meet these demands, Singtel designed Paragon™, a software-defined network platform enabling customers to request quality of service, bandwidth guarantees, and network slices through simple templates. It is the industry’s first multi-network, multi-cloud (edge cloud, public cloud, and AI cloud) orchestration platform that enables enterprises to host mission-critical use cases that require assured network QoS seamlessly.
“We built Paragon to unlock the network’s strength and expose it as a service,” Kumar said. “Customers can configure QoS in minutes, not months.”
Transforming the network required significant behind-the-scenes engineering. Singtel developed a new service orchestrator and connected its AI cloud directly to the 5G packet core, placing GPU resources “one hop away” from edge workloads. The result: latency in the 5–7 ms range for mobile use cases such as drone surveillance, industrial robotics and real-time video analytics.
With Singapore’s nationwide coverage, Singtel can position AI compute in multiple regional exchanges, dynamically adapting to workload movement. “For physical AI, there’s no room for delay,” Kumar said. “You need the right compute in the right place at the right time.”
The age of physical and embodied AI
Kumar cited Jensen Huang’s concept of physical AI, where robots, sensors and autonomous systems become pervasive as the inflection point that will most strain Asia’s infrastructure.
“Soon everything will be AI-enabled: robots, appliances, even everyday devices,” he said. “This will put enormous pressure on the network and the cloud.”
To manage this surge, Singtel is developing a hybrid AI architecture that blends:
- Small language models (SLMs) are deployed on devices and edge nodes for ultra-fast inferencing.
- Large language models (LLMs) in the cloud for more complex reasoning.
- Intelligent sampling to determine which queries stay local and which require cloud processing.
For industries like healthcare and FinTech, Kumar emphasised the rising need for vertical-specific SLMs trained on specialised datasets. Critical-care monitoring, fraud detection and financial compliance cannot rely on generic models.
Creating a regional and global API fabric
No discussion at ITW would be complete without touching on the API economy, and Singtel’s strategy here is as comprehensive as its infrastructure play.
Kumar described a three-layered API monetisation framework:
- Domestic APIs
Singtel has launched services like SingVerify, providing SIM-based authentication, SIM-swap detection, and identity-as-a-service for verticals such as finance and transport. Singtel is also aggregating APIs from other local operators to provide unified national coverage.
- Regional APIs via Bridge Alliance
As a founding member of Bridge Alliance, Singtel contributes to a regional API exchange covering more than 30 operators. Notably, Singtel’s Paragon underpins the technology behind this regional aggregator.
- Global APIs through Aduna
In a significant new step, Singtel and nine global carriers, including AT&T, Orange and Telefónica, have established Aduna, a global joint venture to create a worldwide TM-Forum aligned API exchange.
“Aduna will allow enterprises and OTTs to access network APIs across Asia, Europe, the Middle East and the US through a single interface,” Kumar explained. Full commercial readiness is expected next year, with interoperability between Aduna and Bridge Alliance enabling coverage across 50+ markets.
Moderator Roberts pressed whether Singtel could already support customers looking to deploy AI-enabled drone applications globally. Kumar’s answer: “Soon. Integration is underway, and we expect true global federation within the next year.”
Hyperscalers: Competition or cooperation?
With Singtel’s expansion into AI cloud and sovereign AI, Roberts asked the question many carriers are grappling with: Is Singtel now competing with hyperscalers? Kumar, though, rejected the notion.
“This is not competition, it’s complementarity,” he said. “We’re filling a gap hyperscalers cannot yet serve, especially around sovereign AI. But globally, we partner with them. Our platform lets customers choose between Singtel RE:AI’s sovereign AI cloud platform and hyperscaler clouds depending on the market.”
Singtel continues to host hyperscaler workloads in its hyperconnected AI-ready data centres and relies on them for its internal IT infrastructure.
Predictions: What the next five years hold
Kumar closed with six predictions that drew attention in the room:
- Inferencing demand will dwarf training demand
- As foundational models stabilise, Asia will require orders of magnitude more inferencing chips.
- Agentic AI will reshape enterprises
- Robots, digital agents, and automation will work across departments, finance to supply chain, freeing human teams for higher-value innovation.
- Physical AI will enter the mainstream
- AI-embedded devices will become ubiquitous, driving massive low-latency compute requirements at the edge.
From carrier to AI infrastructure partner for sovereign and enterprise needs
The fireside chat made one thing clear: Singtel is now defined by its connectivity, digital services, and digital infrastructure – purpose-built to serve sovereign, enterprise, and regional AI requirements – and is already reshaping Asia’s digital infrastructure landscape.
As Roberts concluded: “This is a truly plug-and-play digital fabric, network, cloud, data centre and intelligence integrated end-to-end.”
For Asia’s next decade of digital growth, that fabric may prove essential.
RELATED STORIES
Singtel trims Bharti Airtel stake in $1.2bn deal as it sharpens capital strategy
CMI and Singtel partner on enterprise connectivity in Asia Pacific
Datacloud Energy 2026
After a standout 2025 edition, we’re back with an even sharper focus on the intersection of data centres, energy, and ESG. As power demand rises and regulations evolve, there’s a growing urgency to rethink how infrastructure is powered, financed, and built for long-term impact.
