HKT’s announcement of Hong Kong’s first 3.2Tbps AI Data Centre Inter-connect (DCI) Superhighway reflects a global trend in which leading telecom operators are transforming their fibre and optical transport networks into AI-ready infrastructure capable of supporting hyperscale data centres and distributed GPU clusters.
As artificial intelligence workloads become increasingly compute-intensive, operators are investing in ultra-low latency, high-capacity connectivity that allows AI models to be trained and deployed across multiple geographically distributed data centres.
The latest Omdia report revealed telecom operators are investing in AI infrastructure, including AI data centres, GPU-as-a-Service, AI cloud platforms, optical transport, and data centre interconnect (DCI). Operators such as SK Telecom, SoftBank, Singtel, Ooredoo, STC, Iliad, and others reposition themselves as AI infrastructure providers rather than traditional connectivity companies.
IDC report forecasts AI and generative AI spending in telecommunications across EMEA to grow at a 31.8 percent CAGR between 2024 and 2029, with operators investing in AI-ready infrastructure to support new enterprise services and operational transformation.
Several global telecom operators have recently unveiled similar initiatives.
Lumen Technologies has expanded its AI-ready network by deploying dedicated fibre routes and high-capacity wavelength services that connect AI data centres and hyperscale cloud providers, enabling low-latency communication between large GPU clusters.
NTT Communications is advancing AI networking through its Innovative Optical and Wireless Network (IOWN) initiative, using all-photonics technology to deliver ultra-low latency and high-bandwidth connections for next-generation AI infrastructure.
European operators are also strengthening their AI connectivity capabilities.
Sparkle has expanded its AI-ready international backbone with high-capacity optical transport linking major European and Mediterranean data centres.
Orange Business is enhancing its data centre networking portfolio with high-capacity Ethernet and optical services designed for enterprise AI applications.
In the Asia-Pacific region, Tata Communications has integrated AI-ready connectivity into its Digital Fabric platform, combining global networking with AI cloud infrastructure and GPU services to support enterprise AI deployments across multiple regions.
HKT’s latest initiative distinguishes itself by combining a 3.2Tbps data centre interconnect network with advanced hollow core fibre, which is expected to deliver around 30 percent lower latency than conventional fibre networks.
The company is also supporting AI-native technologies including Remote Direct Memory Access (RDMA), RDMA over Converged Ethernet (RoCEv2), Multipath Reliable Connection (MRC) and Ultra-Ethernet Transport, allowing thousands of GPUs across multiple data centres to operate in synchronisation with minimal transmission delays.
The first phase of the network, linking Lok Ma Chau Loop and Tseung Kwan O, is scheduled for completion by the end of 2026 and will support the upcoming Sandy Ridge Data Facility Cluster, reinforcing Hong Kong’s ambition to become a leading regional hub for AI computing and digital infrastructure.
The growing investments by HKT and other global telecom operators underline how AI is reshaping telecom infrastructure strategies. Instead of focusing solely on traditional connectivity services, operators are increasingly building specialised AI transport networks that deliver multi-terabit bandwidth, ultra-low latency and energy-efficient connectivity, creating the digital foundation required for the next generation of AI applications, cloud computing and high-performance data centres.
BABURAJAN KIZHAKEDATH
