Scenarios

Chip and AI, China accelerates and aims to triple production by 2026

China's most powerful innovation players take the field: from Huawei to Deep Seek

by Biagio Simonetta

3' min read

3' min read

Bans and tariffs are not enough. The Chinese innovation agenda does not change and does not stop. This seems to be what Beijing's semiconductor plan is all about. Because if the United States continues to press the stop pedal, with the aim of halting the Chinese race for artificial intelligence, the response is anything but submissive.

According to the Financial Times, the country is preparing for an unprecedented acceleration in the production of chips for artificial intelligence. And it is doing so with two clear objectives in mind: to close the gap with the United States and to reduce dependence on foreign suppliers (see under Nvidia).

Loading...

Beijing aims to triple the national capacity of AI processors as early as 2026, a goal that passes through new production facilities and increasing integration between hardware manufacturers and software developers.

The long vicissitudes of the last three years, which have seen Nvidia play a leading role (with its depowered processor, H20), have convinced the Chinese government that the road of technological autonomy is the only possible way to play the leading role. Hence investments and strategic moves.

The most powerful players have been called upon to take the field. Starting with Huawei, which will open three factories dedicated to chips between the end of 2025 and the beginning of 2026, And the combined capacity of these factories, once fully operational, could exceed the current capacity of SMIC, China's leading foundry at the moment. SMIC itself plans to double its 7-nanometer chip production in 2026, the highest level of technology available in China today. Huawei is the main customer for these lines, but the expansion will also open up space for smaller players such as Cambricon, MetaX and Biren, who will be able to count on more resources to challenge a market left bare by Nvidia after the export restrictions imposed by Washington.

But a leading role in this Chinese race will also be played by DeepSeek, a start-up that has become famous in recent months for emulating the power of Western chatbots (such as ChatGPT) using much more limited resources. DeepSeek has recently defined a new technical standard based on the FP8 data format. Huawei and Cambricon already have compatible products, while other players are developing similar solutions.

It is precisely the adoption of this standard that could enable Chinese companies to improve the efficiency of domestic chips, in what would be a real turning point. Although the alignment of hardware and software, Nvidia's winning key, may require years of collaboration and investment not only in processors but also in complementary areas such as memory and connectivity. And it is no coincidence that the Chinese government is also pushing CXMT (leading domestic memory manufacturer), which is testing samples of HBM3, a technology only a generation behind the more advanced solutions used by Nvidia GPUs.

Beijing's political and financial support is evident. The State Council has called for accelerated adoption of AI and greater integration between research, development and commercialisation. The operation is ambitious: to create a cohesive ecosystem between chip manufacturers, memory suppliers and AI model developers, along the lines of what has happened in the West around Nvidia. As DeepSeek founder Liang Wenfeng pointed out, 'Nvidia's advantage does not come from a single company, but from a collective effort of an entire industry. China needs the same ecosystem to build its leadership'. And Beijing seems to have taken this literally.

Copyright reserved ©
Loading...

Brand connect

Loading...

Newsletter

Notizie e approfondimenti sugli avvenimenti politici, economici e finanziari.

Iscriviti