Chinese chipmakers aim to triple the country's total AI processor production next year. This is due to Beijing competing with the United States in developing state-of-the-art artificial intelligence, writes UNN with reference to the Financial Times.
Details
One factory, designated to produce Huawei AI processors, is set to begin production later this year, with two more to be launched next year, two sources familiar with the matter said.
While the new factories are specifically designed to support Huawei, it is unclear who exactly owns them. Huawei denied plans to launch its own factories and did not provide further details.
Chinese companies are also competing to develop the next generation of AI chips tailored to a standard supported by DeepSeek, which has become a leading AI startup in the country. Huawei's latest products are seen as among those that will meet DeepSeek's requirements.
The combined capacity of these three new factories, once fully operational, could exceed the current total output of similar lines at Semiconductor Manufacturing International Corporation (SMIC), China's leading fab, according to sources familiar with the efforts.
According to sources, SMIC also plans to double its 7-nanometer chip production capacity next year — the most advanced mass-produced type in China. Huawei is currently SMIC's largest customer for such processor lines.
As a result, smaller Chinese chip developers such as Cambricon, MetaX, and Biren will be able to gain much greater SMIC capacity, stimulating competition in the rapidly growing Chinese market left by Nvidia after the US export ban.
Semiconductors are at the heart of trade tensions between Beijing and Washington. The US has restricted access to leading AI processors produced by global leader Nvidia in an attempt to curb China's efforts to develop artificial intelligence.
Domestic production will not be a problem for long, especially considering that all capacities will be launched next year
Chinese companies are also competing to develop the next generation of AI chips tailored to a standard supported by DeepSeek, which has become a leading AI startup in the country.
Last week, DeepSeek announced that its models now use the FP8 data format, designed to adapt to the next generation of domestic chips, without specifying from which supplier.
Share prices of listed Chinese semiconductor companies such as Cambricon and SMIC soared after the announcement.
Huawei's 910D, as well as Cambricon's 690, are considered leading products designed to support Deepseek's desired standards, while several smaller Chinese chipmakers are also accelerating the development of their versions.
DeepSeek chose this data format because it increases hardware efficiency, but at the expense of accuracy, which can help Chinese AI players compete with international competitors, even if their chips are a generation behind Nvidia's leading offerings.
If we can develop and optimize these Chinese chips to train and run Chinese models in the ever-changing Chinese ecosystem, one day we will remember this shift as an even more significant DeepSeek moment. This unprecedented alignment can compensate for our less developed hardware capabilities
However, such efforts will require years of cooperation between computing chip manufacturers and those who develop memory and connectivity hardware, as well as the software tools needed to support them.
Addition
Nvidia CEO Jensen Huang arrived in Taipei to meet with TSMC, discussing the release of new chips and the situation with H20. The company is working on a successor to H20, but the decision on its supply to China depends on the US government.