sdecoret - stock.adobe.com
How APAC telcos are reclaiming relevance with AI-ready infrastructure
In part three of this series, we explore how Singtel, China Telecom and SK Telecom are pivoting away from head-on competition with hyperscalers to become enablers of the AI era through sovereign cloud and edge platforms
In parts one and two of this series, we explored how hyperscalers and systems integrators are squeezing telcos and capturing enterprise value by translating cloud capabilities into verticalised outcomes. Now, we examine how Asia-Pacific (APAC) telcos can reclaim relevance through artificial intelligence (AI)-ready infrastructure.
Across the region, leading telcos are showing that success in the AI era doesn’t come from competing head-on with hyperscalers. It comes from building complementary infrastructure like AI-ready datacentres, sovereign clouds, and edge platforms. These enable enterprise transformation without competing head-on.
Here’s a look at what telcos across the region have done:
Nxera: Singtel’s infrastructure as enabler strategy
Singtel’s Nxera is more than a datacentre operator. Designed for AI workloads, it offers low-latency, high-density compute environments for both hyperscaler and enterprise clients across Asia-Pacific.
Paired with NCS’s enterprise services, Singtel is able to participate in high-value AI initiatives. Nxera turns infrastructure into momentum, helping enterprises scale AI faster, smarter, and more sustainably.
Singtel’s dual-path strategy allows it to enable transformation without directly competing with hyperscalers.
Xirang: China Telecom’s AI embedded network
China Telecom is redefining what telco infrastructure can be. Its Xirang platform is one of the boldest examples of telco-led AI infrastructure. It hosts a 177-billion-parameter large language model distributed across 500 kilometres of telecom infrastructure, surpassing GPT-3.5 in scale.
Powered by 1,024 graphics processing units (GPUs), Xirang enables smart factories, cities, and industrial automation, all while maintaining data sovereignty. It’s also part of China’s national supercomputing internet, turning AI into a distributed utility.
Today, Xirang is being used by a manufacturing firm to reduce model training costs and speed up quality control while a biotech startup is using it in molecular simulations, which can now be completed overnight compared to three months previously.
By integrating AI into the network itself, China Telecom has shown that telcos can provide enterprise capabilities that hyperscalers cannot easily replicate, particularly for latency sensitive, sovereign workloads.
SK Telecom: Sovereign AI cloud and GPU-as-a-Service
South Korea’s SK Telecom is taking a differentiated route – building a sovereign AI cloud and GPU-as-a-service (GPUaaS) to enterprises. This allows SKT to retain control over critical workloads while enabling scalable AI adoption across verticals such as manufacturing, finance, and smart cities.
By owning the infrastructure stack, SKT ensures compliance, performance, and strategic relevance in a hyperscaler-dominated market. Its sovereign cloud aligns with national data policies, while GPUaaS provides flexible access to high-performance compute. SKT’s model shows how telcos can monetise infrastructure without competing directly while becoming indispensable enablers of AI transformation.
Telstra: A strategic reflection
Australia’s Telstra made early moves to build its own cloud capabilities – a bold step that reflected the ambition of many telcos to compete in a hyperscaler-dominated landscape. However, like others in the industry, it faced challenges in matching the scale, economics, and pace of innovation that global cloud providers could sustain.
Telstra’s eventual pivot toward partnerships illustrates a broader lesson for APAC telcos: success in the AI era may depend less on competing directly, and more on enabling transformation through collaboration, infrastructure, and differentiated services.
Three lessons for APAC telcos
Across APAC, telcos are learning that infrastructure isn’t just about hosting workloads. Rather, it is about enabling transformation. Whether through sovereign AI platforms, edge compute, or hyperscaler-friendly datacentres, infrastructure ownership offers telcos a way to stay relevant without being commoditised.
Enable hyperscaler adoption
Build AI-ready datacentres that support cloud-native workloads while offering latency and compliance advantages. Singtel’s Nxera exemplifies this by enabling hyperscaler and enterprise AI workloads through high-density, low-latency infrastructure without competing directly.
Own sovereign AI platforms
Develop infrastructure aligned with national AI strategies and data sovereignty. China Telecom’s Xirang embeds a 177B-parameter LLM across 500km of network, delivering sovereign, low-latency AI services for industries like manufacturing and smart cities.
Monetise edge and GPU services
Offer compute-as-a-service models tailored to vertical needs and regulatory demands. SK Telecom’s sovereign AI cloud and GPUaaS enable scalable AI adoption while maintaining control over critical workloads.
Edwin Lin is principal consultant at Omdia, part of Informa TechTarget
