State-owned telecom giant China Mobile announced plans to use only homegrown chips for the nation’s largest artificial intelligence computing network by 2028, accelerating China’s efforts to become a leader in AI and reduce its reliance on foreign technology.
China Mobile chairman Yang Jie outlined the ambitious target at the company’s Global Partners Conference in Guangzhou on Saturday.
The telco, he said, planned to double its investment in AI and explore the construction of a computing cluster with 100,000 graphics processing units (GPUs), taking its national total to 100 EFLOPS of AI computing power.
An EFLOP – a unit of computing speed equal to one quintillion floating-point operations per second – is used to rate the performance of the world’s most powerful supercomputers.
While he did not specify the precision target of the 100 EFLOPs, using an industry standard of FP16 would mean that China Mobile would effectively be tripling its total AI computing power, which at the end of 2024 stood at 29.2 EFLOPS at FP16.
FP16, also known as half-precision floating-point format, is a computer number format that uses 16 bits (half the standard 32 bits) of memory. Higher precision means that the computing chips process information more accurately, but at a higher cost.
“Human society has fully entered the ‘AI Plus’ era, with AI as the core engine of new productive forces,” said Yang. “Artificial intelligence will not replace humans, but will instead handle repetitive and rule-based work, while humans will continue to lead in emotion, creativity, and moral judgement.”
