Uniquely positioned

But why do industries turn to a telecom operator for their SLMs?

Guo noted that telecoms bring a unique set of strengths to AI, as their role is not just about building models, but about making models that are interoperable and secure.

AI does not work in isolation; it needs to be connected with tools such as 5G, cloud computing, and high-performance data centers, which are areas where telecom operators sit at the intersection of these technologies, Guo said.

For example, in a smart port, China Unicom can deploy the AI models on its cloud platforms and connect AI with internet of things devices through its 5G networks, Guo said.

The cloud acts as a central brain, processing vast amounts of data with AI models, while 5G networks connect hundreds of cameras, sensors, vehicles, and machines operating at once with ultra-low latency.

This highlights another advantage of telecoms in developing AI models: providing end-to-end solutions for its partners, Guo said. Clients can access the services they need through a single provider, rather than juggling multiple vendors.

At the core of this integration lies computing power, a field where China Unicom stands out with its robust capabilities, both Xue and Guo pointed out.

In July, Chen Zhongyue, chairman of China Unicom, unveiled a plan to accelerate the company's intelligent computing infrastructure, including exploring a 100,000-GPU cluster deployment and targeting 45 EFLOPS of computing power by the end of 2025.

EFLOPS is a unit of the speed of computer systems and is equal to 1 quintillion floating-point operations per second.

The plan is expected to position China Unicom among the world's most powerful AI computing benchmarks.