When Singtel’s community went darkish for eight hours on March 16, the ripple results had been speedy and far-reaching. Emergency providers faltered. Digital funds stalled. 1000’s of gig staff misplaced a complete day’s earnings. For a city-state that has staked its financial future on digital management, the outage was greater than an inconvenience: it was a stress take a look at the infrastructure didn’t completely go.
The incident arrives at a pivotal second. Southeast Asian (SEA) telcos are bracing for cellular information consumption to surge to 40GB per consumer by 2030, pushed largely by the accelerating deployment of AI throughout each sector of the financial system. On the coronary heart of this transformation is an idea gaining pressing traction in boardrooms and coverage circles alike: Sovereign AI — the precept that nations should personal, function, and govern the AI infrastructure that underpins their essential programs, relatively than cede that management to international platforms or distant cloud suppliers.
For Singapore, Sovereign AI shouldn’t be merely a geopolitical aspiration. It’s an infrastructural crucial. As AI workloads demand always-on compute, low-latency information processing and ironclad community reliability, the query is whether or not the telco sector — the spine of the digital financial system — is architected for what comes subsequent.
Mayank Srivastava, chief govt of BDx Information Centres, argues the Singtel outage carries a lesson that extends effectively past fault attribution. “As economies digitise, dependencies focus throughout networks, information centres, and cross-border hyperlinks,” he stated in an e mail interview with e27.
Additionally Learn: Singtel launches US$250M AI fund to show its telco empire into an AI deployment platform
The next is an edited excerpt of the dialog.
Are you able to clarify what Sovereign AI means in follow, and why you consider Singapore and SEA dangers “information colonisation” if it doesn’t act now?
In sensible phrases, Sovereign AI is about making certain that worth creation occurs regionally. Within the AI financial system, the information centre is the manufacturing facility housing massive‑scale GPU clusters. It’s the place information is processed, fashions are educated, and choices are generated. Sovereign AI implies that information, compute, and governance frameworks are aligned inside nationwide or regional jurisdictions, below native legal guidelines and accountability.
This isn’t about isolation or exclusion. It’s about financial participation. Traditionally, economies that exported uncooked supplies however imported completed items captured much less lengthy‑time period worth. Within the digital financial system, information is the uncooked materials. Whether it is persistently processed elsewhere, the financial and strategic worth related to it accumulates exterior the area.
The implication isn’t just income, it’s company. Important programs in healthcare, finance, and public providers more and more depend on AI‑pushed resolution layers. Guaranteeing that these programs are supported by trusted, regionally ruled infrastructure strengthens transparency, resilience, and public belief. Once we use the time period “information colonisation,” we’re referring to this financial worth‑seize dynamic, not a political idea.
Singapore is effectively positioned to steer on this space. Current IMDA initiatives round trusted infrastructure, AI governance, and excessive‑effectivity information centres replicate a considerate, ahead‑wanting method. By supporting safe, power‑environment friendly AI infrastructure inside its regulatory framework, Singapore can anchor worth creation regionally whereas remaining globally linked—benefiting enterprises, startups, and the broader digital financial system.
With SEA telcos projecting 40GB of knowledge per consumer by 2030, what does that demand curve truly imply for the bodily infrastructure required to assist it?
The headline quantity has certainly shifted. With 5G-Superior, V2X, and chronic machine-to-machine site visitors, SEA’s cellular information utilization is now projected at round 38-40GB per consumer by 2030 per Ericsson and GSMA baselines, although aggressive AI/IoT situations might push towards 60GB+ in high-growth markets. However the actual infrastructure implication isn’t simply quantity. It’s the form of the demand curve.
Additionally Learn: Echelon Philippines 2025 – Constructing at telco-scale: How startups can leverage Globe’s ecosystem for fast-track market entry
What issues is the character of the information. Extra extremely‑excessive‑definition video, way more actual‑time AI inference, and steady IoT site visitors basically change infrastructure necessities. These are latency‑delicate, at all times‑on workloads that stress energy supply, cooling, and community resilience in methods conventional cellular site visitors by no means did.
Round 40GB per consumer shouldn’t be a delicate enhance; it’s a vertical climb when translated into physics. 5 years in the past, a typical rack ran at 5kW. As we speak, NVIDIA DGX GB200-scale AI racks attain 120-200kW in manufacturing (scaling to 700-800kW in dense clusters). By 2030, 1-2MW per rack is reasonable as energy density turns into the restrict.
Supporting 2MW racks requires an order‑of‑magnitude shift in cooling, together with direct‑to‑chip and liquid‑immersion programs, together with redesigned energy trains and grid interfaces. As highlighted in BDx discussions on AI‑first services, this represents 10× cooling retrofits in comparison with standard designs.
Regionally, this demand can’t be met by any single market alone. It factors to the necessity for coordinated capability improvement throughout Singapore, Indonesia, Malaysia, Thailand, and Vietnam. Given the lengthy lead instances concerned in energy provisioning and development, infrastructure planning should transfer effectively forward of demand relatively than react to it.
Do you assume telcos ought to cease worrying about energy density and refocus on providers that drive development? What’s the hazard of telcos persevering with to construct and function their very own information infrastructure relatively than partnering with specialist suppliers like BDx?
Telcos face a real strategic balancing act. Trendy information centres have advanced into extremely specialised environments requiring deep experience in energy engineering, thermal administration, and more and more AI‑optimised design. These capabilities sit alongside—however are distinct from—core telecommunications operations.
The query is much less about functionality and extra about focus. As networks evolve and providers grow to be extra subtle, tying up capital and management consideration in extremely specialised infrastructure can restrict flexibility elsewhere.
Additionally Learn: A brand new daybreak within the post-2G period: How cloud know-how can propel the telco trade to new heights
Partnership fashions supply an alternate. By working with specialist suppliers, telcos can entry AI‑prepared, future‑proof infrastructure whereas concentrating their funding and innovation efforts on community high quality, platforms, and buyer‑going through providers. This separation of roles additionally lowers friction for startups and enterprises, who profit when telcos give attention to service innovation whereas infrastructure specialists give attention to scale, effectivity, and resilience.
It’s a collaborative method that enables every participant to function the place they add probably the most worth.
How ought to we take into consideration the ROI of reliability? When a telco goes down, the prices are apparent, however what’s the enterprise case for investing closely in resilience earlier than a disaster occurs, particularly when margins are already below strain?
Reliability investments are difficult to justify as a result of their worth is most seen in what doesn’t occur—outages averted, prospects retained, regulatory scrutiny prevented. Conventional ROI fashions wrestle to seize this.
A helpful analogy is healthcare. There may be an accepted baseline of reliability under which programs merely can not function. In a digital financial system, communications infrastructure more and more occupies that very same class. As AI helps actual‑time finance, healthcare, and public providers, reliability turns into a prerequisite relatively than a differentiator.
In that context, resilience shouldn’t be defensive spending. It’s a situation for collaborating in increased‑worth use instances. Operators that may show constant, measurable reliability function in a special industrial and regulatory dialog than these competing solely on value.
As Southeast Asian telcos consolidate to spice up valuation, there’s a pressure between leaning out operationally and constructing the strong spine an AI-native financial system wants. How do telcos resolve that contradiction?
Consolidation can create scale, however scale alone doesn’t remedy architectural complexity. The chance lies in being exact about what to optimise internally and what to entry by means of partnerships.
Additionally Learn: Founders’ playbook for resilience in 2026: Constructing in atoms in a fractured world
AI goes to be a utility, like electrical energy or the web. The spine required for an AI‑native financial system is layered. It consists of networks, specialist infrastructure, cloud platforms, and regulatory frameworks working collectively. No single steadiness sheet must personal each layer.
By treating infrastructure as one thing to entry strategically relatively than personal completely, telcos can redirect capital towards community high quality and differentiated providers whereas nonetheless supporting the depth and resilience AI workloads require.
Should you had been advising Singapore’s policymakers as we speak, what are the 2 or three most pressing infrastructure choices they should make within the subsequent 12 to 24 months to make sure the nation is genuinely prepared for an AI-driven future?
First, aligning energy coverage with AI timelines.
AI infrastructure investments transfer quicker than conventional approval cycles. Whereas Singapore’s regulatory rigour is a energy, there’s scope for clearer, quicker pathways for top‑effectivity, AI‑optimised capability with outlined sustainability requirements. As AI adoption accelerates, the subsequent 12 to 24 months grow to be disproportionately essential for setting these frameworks.
Second, strengthening trusted compute for essential sectors.
As AI turns into integral to finance, healthcare, and public providers, making certain that these workloads are supported by resilient, trusted infrastructure is crucial. Periodic stress‑testing of dependencies and inspiring significant infrastructure variety can additional strengthen confidence.
Third, conserving rules sensible and enabling.
Singapore has a robust observe report of utilizing regulation to unlock innovation relatively than constrain it—throughout sectors reminiscent of aviation, fintech, and border safety. Changi Airport’s use of facial recognition is a transparent instance of regulation offering the readability and confidence wanted for giant‑scale adoption.
Additionally Learn: A brand new daybreak within the post-2G period: How cloud know-how can propel the telco trade to new heights
AI infrastructure requires the identical seamless method: clear guidelines, aligned incentives, and powerful governance, applied in a approach that matches the tempo and capital depth of the know-how. When regulatory frameworks are predictable and outcomes‑based mostly, they permit infrastructure suppliers to take the lengthy‑time period funding dangers required to maintain Singapore—and the area—on the forefront of world AI improvement.
This steadiness between oversight and enablement is one in all Singapore’s defining strengths, and making use of it thoughtfully to AI infrastructure shall be key to sustaining management because the ecosystem continues to evolve.
—
Picture Credit score: Taylor Vick on Unsplash
The submit When the spine breaks: Can Singapore’s telcos energy a Sovereign AI future? appeared first on e27.
