Rising AI demand is forcing Asia-Pacific telecoms to upgrade networks for lower latency, higher bandwidth and tighter cross-border coordination. Rising AI demand is forcing Asia-Pacific telecoms to upgrade networks for lower latency, higher bandwidth and tighter cross-border coordination. Read More DataCenterNews Asia Pacific
GSMA convened an Asia-Pacific roundtable on computing-network integration for the AI economy in Shanghai, bringing together government officials, telecom operators, technology companies and standards bodies from across the region.
Discussions focused on how networks in Asia-Pacific must evolve as demand for AI services grows. Participants highlighted the need for lower latency, stronger reliability, greater bandwidth and closer coordination of cross-border infrastructure, computing resources and standards.
Executives and officials described a shift from broad connectivity to systems built for AI training, inference and industrial use. That requires tighter links between computing centres, backbone networks and access networks, along with better international interconnection across submarine and land cables.
In opening remarks, Si Han Bo CHEN, Head of GSMA Greater China, said existing approaches were no longer sufficient for the next stage of AI deployment. She added that the region had moved from being a technology follower to taking a leading role in new infrastructure and computing-network integration.
China used the gathering to outline its planning model for computing and network integration. He Xiaowen, Deputy Director-General of the Department of Planning at China's Ministry of Industry and Information Technology, said the country is pursuing a three-tier structure covering national, regional and urban layers, alongside high-speed direct links between regions.
He said China is advancing all-optical networks, 400G and 800G transmission, SRv6 and intelligent computing orchestration to reduce what he described as computing silos. He also pointed to a "millisecond" computing action aimed at achieving millisecond-level interconnection between computing centres, access to computing resources and delivery of computing applications.
On regional coordination, He called for stronger interoperability and resource sharing across Asia-Pacific. "Computing power empowers the future, networks connect us all, and synergy achieves win-win results," he said.
Standards push
A recurring theme was the role of international standards in making AI networks interoperable and reliable. Dr Atsuko Okuda, Regional Director of the International Telecommunication Union Regional Office for Asia and the Pacific, said telecom operators across the region are likely to face major infrastructure upgrades as AI drives demand for data and computing.
She pointed to fibre expansion, backbone upgrades and denser 5G deployments. Okuda also said countries should refer to the more than 200 AI-related standards developed at the ITU when deploying AI systems, while ensuring less-connected communities are not excluded from digital change.
Thailand and Sri Lanka both framed the issue as part of national economic policy. Chaichanok Chidchob, Minister of Digital Economy and Society of Thailand, said resilient digital and mobile connectivity underpins the AI economy and linked the issue directly to the country's development strategy.
Eng. Eranga Weeraratne, Deputy Minister of Digital Economy of Sri Lanka, said smaller economies must decide whether they will only consume AI or also produce it in selected areas. He said Sri Lanka had chosen a selective production route, supported by regional partnerships and closer links between computing and networks.
Operator role
Several telecom and cloud groups said operators will play a larger role in AI deployment than simply carrying traffic. Wu Jiong, Marketing Deputy General Manager at Mobile Cloud, said China Mobile is building Intelligent Computing Centres and all-optical backbone infrastructure as part of a broader plan to coordinate cloud, edge and terminal resources.
He said the company aims to improve access to different types of computing resources through scheduling and optimisation. "In the AI era, security is not an option, it is a mandatory question," Wu said.
Huawei used the roundtable to argue that the region needs a common technical direction. Zhai Haipeng, Vice President of Huawei's Optical Business Product Line, said AI is pushing networks toward latency of 5ms to 1ms, reliability of 99.999% or 99.9999%, and greater bandwidth.
He called for alignment around the ITU-T's ION-2030 architecture and identified three priorities for Asia-Pacific: coordination of cross-border submarine and land cables, deterministic low-latency networks built around 1-5-20ms latency circles, and an ultra-gigabit strategy.
AI model developers also argued that operators could become commercial partners rather than just infrastructure providers. Daniel Shi, Global Business Account Executive at MiniMax, said deeper cooperation could include resource exchange and wholesale distribution arrangements as AI demand rises.
He said future networks for AI services would need Tbps-level bandwidth, millisecond-level latency and strong security. "The black magic of the AI field must be held in the hands of the good guys," Shi said.
Commercial pressure
Kuaishou and CTM Macau offered examples of how these network issues are reaching end-user markets. Liu Zhen, Vice President of Kuaishou Group, linked low-latency, reliable computing-network infrastructure to commercial success in short video, eCommerce and AI services, while arguing that business decisions should be closely tied to user demand.
CTM Macau said smaller markets can use local network strength to support AI adoption. Ebel Cham, VP Commercial at CTM Macau, said telecom services remain the foundation of AI development and outlined the company's effort to create an "AI Hub" by aggregating models, computing resources and applications through common billing and platform interfaces.
Another commercial argument came from Zhang Chao, Vice President of Kunzhen Zhenyu and President of the Overseas Business Division, who linked computing resources directly to revenue generation in the AI market. "Computing power is directly equal to revenue, and tokens have become the new commodity of the AI era," Zhang said.