China’s AI Landscape: China Mobile’s Ambitious GPU Cluster Initiative
Summary:
- China Mobile is set to develop a 100,000-card GPU cluster by 2028, leveraging only domestic chips.
- This initiative aims to enhance AI autonomy and decrease reliance on foreign technology.
- The company’s capital expenditures for 2024 will reach approximately $22.99 billion, highlighting its commitment to AI infrastructure.
Nvidia’s dominance in the Chinese AI chip market has dramatically declined, dropping from a staggering 95% market share to virtually none. This seismic shift underscores a critical juncture in the technological landscape of China, where the focus is pivoting towards self-sufficiency.
China Mobile’s Strategic Plan
On October 18, 2023, China Mobile unveiled a groundbreaking initiative to establish a massive AI computing network that will exclusively utilize domestic chips by 2028. This strategic move aligns with China’s broader goal of enhancing artificial intelligence independence amid ongoing geopolitical tensions.
According to Yang Jie, Chairman of China Mobile, this initiative represents an advanced version of the "AI+" action plan. The aim is to create a robust national AI computing cluster grounded solely in domestic technology. This will not only accelerate the country’s ambitions in AI but also fortify its technological sovereignty.
Ambitious Goals for 2028
China Mobile plans to construct a GPU cluster consisting of 100,000 cards, targeting over 100 exaFLOPS in computational power. This ambitious project is indicative of the company’s commitment to not just compete, but to lead in the AI sector.
The company’s recent annual report highlights a capital expenditure of approximately 164 billion yuan (around $22.99 billion) for 2024, accounting for 18.4% of its total revenue. This significant financial investment is earmarked to develop AI capabilities that will position China Mobile as a formidable player in the AI value chain.
Growing Computing Capacity
By the end of 2024, China Mobile aims to achieve an intelligent computing capacity of 29.2 exaFLOPS using FP16 accuracy. This figure is projected to expand to 33.3 exaFLOPS in the first half of 2025. By incorporating rented resources, the total capacity could soar to 61.3 exaFLOPS, further solidifying its computing power.
The planned investment of 37.3 billion yuan in computing power in 2025 will constitute 25% of the company’s capital expenditures. It aims for a total computational scale (FP32) of 8.9 exaFLOPS, with intelligent computing (FP16) surpassing 34 exaFLOPS.
A Departure from Traditional Models
Unlike many competitors who typically lease AI hardware, China Mobile is committed to developing independent capabilities through its proprietary "Nine Days" large-scale AI model. This shift allows the company to evolve from merely selling computing resources to offering robust AI capabilities. As a result, China Mobile is poised to play a more significant role in the AI industry’s value chain, lessening its dependency on global AI providers.
Shifting Market Dynamics
A report from Epoch AI indicates that as of May 2025, the U.S. controls approximately 75% of the world’s AI computing power, while China commands about 15%. In this context, China Mobile’s thrust towards building a self-sufficient AI infrastructure is not just strategic but essential for closing this gap.
Conclusion
China Mobile’s ambitious plans for a 100,000-card GPU cluster mark a transformative phase in the Chinese AI landscape. The company’s substantial investments and focus on domestic technology pave the way for an era of increased autonomy in artificial intelligence. As the global technology landscape evolves, initiatives like these will play a crucial role in shaping the future of AI and ensuring that countries can stand on their own in this highly competitive field.
With a forward-thinking approach that emphasizes self-reliance, China Mobile is not just preparing for the future; it is actively reshaping the dynamics of international technology competition.
