On April 10, at the main forum of the 2025 Mobile Cloud Intelligence Computing Conference, Yang Jie, chairman of China Mobile, delivered a speech and proposed the strategic path "from cloud to intelligence". He pointed out that the core carrier of the computing power network is upgrading from the traditional "cloud computing" to the deep integration of "cloud and AI" known as "cloud intelligence computing".

Yang Jie stated that China Mobile has innovatively proposed the concept of the "computing power network", promoting three directions: "diversified computing power", "integrated computing and network", and "comprehensive AI". Currently, more than one million computing servers have been deployed, 13 national and regional intelligent computing center nodes have been built, multiple super-large intelligent computing centers have been created, with an intelligent computing scale exceeding 43 EFLOPS. A total of 21 intelligent computing centers, 3 national-level supercomputing centers, and 3 quantum computing centers are integrated into the network, and the schedulable computing resources account for one-sixth of the country's total.

He revealed that China Mobile has taken the lead in completing the 400G high-speed direct connection network of the eight national hubs. The computing network brain has been implemented in regions such as Beijing-Tianjin-Hebei and the Yangtze River Delta. In the Wuhu cluster of the Yangtze River Delta, China Mobile has built the first "four-in-one" scheduling platform integrating general computing power, intelligent computing power, supercomputing, and quantum computing.

Internally, China Mobile has constructed an AI-driven computing and networking fusion scheduling platform, supporting over 100 million cross-regional computing and networking dispatches per day. Externally, 24 products, including the "Linxu" intelligent body, have undergone AI upgrades, with an AI user base exceeding 190 million. The company has also launched new services for AI model training and inference, implementing over 40,000 scenario-based solutions.

Yang Jie pointed out that the "scale effect", "rebound effect" of AI, and the "acceleration effect" of connectivity are driving profound changes in the computing network architecture. By 2030, AI tasks will account for 64% of all network traffic, the demand for inferencing computing power will surpass training computing power, and the number of robots and intelligent bodies will exceed the human population.

To adapt to the needs of AI development, China Mobile will expand its traditional IaaS-PaaS-SaaS system into four layers: AI IaaS, AI PaaS, MaaS, and AI SaaS, covering the entire chain from computing resources to AI applications, strengthening infrastructure, tool platforms, model services, and native product supply capabilities.

In building a converged computing network infrastructure (AI IaaS), China Mobile plans to construct a super-large "computing factory", promoting the doubling and expansion of smart computing centers at the ten-thousand-card and thousand-card levels. It will also conduct research on ten-thousand-card smart computing centers, build "centralized + distributed" inferencing computing resources, and form an intelligent computing system featuring "centralized concentration, edge ubiquity, centralized training and decentralized inference, and integrated training and inference".

At the same time, it will build a "network hub" with a three-dimensional layout. Accelerate the application of 6G characteristics such as integrated天地一体, intelligent fusion, and communication-sensing integration in current networks, strengthen the pilot implementation of 800G and 1.2T ultra-high-speed next-generation networks, and expand low-altitude networks, aviation networks, and satellite networks to support "human-machine-thing"全域connection.

In the future, China Mobile will focus on four key directions: "converged computing network infrastructure", "self-service tool platform", "one-stop model service", and "native product application", promoting the formation of an AI full-stack service system covering "computing power-tool-model-application", and creating a service matrix consisting of intelligent cloud, intelligent terminals, and intelligent bodies.

Original article: https://www.toutiao.com/article/7491819019985617460/

Disclaimer: This article represents the author's personal views. Please express your opinions by clicking the "upvote" or "downvote" buttons below.