Published on Dec 17, 2025

MetaX’s share price rose by roughly 600 to 700 percent on Wednesday, the day of its stock-market debut in Shanghai.

Like its rival Moore Threads, the Chinese AI chipmaker is benefiting from the slightly premature expectation that China will one day need far fewer semiconductors from Nvidia and AMD. That transition will still take some time, but many observers regard it as a plausible long-term development.

“Local substitution” — China’s stated goal of building domestic supply chains for GPUs and other chips — is, regardless of timelines and realistic forecasts, one of the main drivers behind the sharp price surge at the IPO.

At one point, within hours of the listing, MetaX shares climbed from 104.66 yuan to 835 yuan, an increase of 697 percent.

Investor enthusiasm also reflects the assessment that the Chinese company has significant growth potential in the coming years, regardless of how relations between the United States and China evolve.

For an AI chip manufacturer, there is hardly a better moment to go public than at the peak of a new wave of AI enthusiasm. Nationality and geopolitical rivalries between governments play only a secondary role in such phases.

China’s semiconductor market is, in any case, so large and is growing so rapidly due to the fast integration of artificial intelligence across virtually all industries that there is ample room for Nvidia and other manufacturers.

That is certainly true for newcomers such as MetaX, which, according to a report by the news agency Reuters currently holds a market share of around one percent in China’s AI chip market.

Together with Moore Threads, Biren Technology, and Enflame, MetaX belongs to China’s “four small GPU dragons”. Moore Threads, whose IPO earlier this month was accompanied by a share-price increase of more than 400 percent, is also sometimes referred to as the “little Nvidia”.

That comparison is likely to make Moore Threads’ management uneasy, because Nvidia and its CEO Jensen Huang are not only highly successful but have also become caught up in the geopolitical tensions between Washington and Beijing.

The U.S. government has banned the sale of Nvidia’s most powerful AI chips to China. On December 8, however, US President Donald Trump gave the green light for exports of the H200, Nvidia’s second-fastest AI chip, to China.

Shortly thereafter, China’s Ministry of Industry and Information Technology convened a crisis meeting in Beijing, according to the technology portal Kuai Keji. In addition to officials from the powerful National Development and Reform Commission, the country’s top planning body, representatives of major AI companies including Alibaba, ByteDance, and Tencent were invited.

What exactly was discussed is not known. Since then, however, rumours have circulated that the Chinese government may not grant import approvals for the H200 after all. A more likely, though still speculative, interpretation is that Beijing has urged major data-centre operators to purchase domestically produced AI chips wherever possible rather than relying on U.S. products. Chinese commentators described Nvidia’s chips as “sugar-coated bullets” (tang yi pao dan), a phrase Mao Zedong once used to describe the seductive tactics of hostile “capitalists”. The concern is that continued access to Nvidia chips could cause China to lose sight of the urgency of reducing its dependence on Nvidia and AMD.

That scenario, however, seems unlikely. China began pursuing its strategy of local substitution (guochan tidai) as early as 2014, long before US restrictions on advanced chips, software, and semiconductor manufacturing equipment. The “chip war” has merely intensified these efforts.

Huawei has already developed AI chips that, while not yet fully matching the performance of Nvidia’s H200 — let alone the Blackwell chip — are no longer far behind. With Huawei’s Ascend 910C, virtually all tasks can now be performed for which Nvidia chips would previously have been the default choice.

Huawei is not yet able to supply the volumes needed to fully meet demand in the Chinese market, but the company has announced plans to at least double its capacity by next year.

While Chinese manufacturers are still catching up in the most advanced AI chips, the picture looks very different in other segments. In automotive and memory chips, as well as in legacy chips at 22 nanometres and above, China’s substitution policy has already made substantial progress. Even in automotive chips there are still niches — such as safety-critical components — where the market is almost entirely dominated by European and Japanese suppliers, but these niches are steadily shrinking.

At the same time, new trends are emerging in China’s AI and semiconductor markets that could soon render many U.S.–China comparisons obsolete. One of these is the growing reliance of Chinese companies on open-source solutions, both for large AI models — with Alibaba as a prominent example — and for chip architectures such as RISC-V.

The traditional “moats” that long secured strong market positions for foreign chipmakers in China, such as Nvidia’s CUDA ecosystem, are eroding ever more quickly. If Chinese AI companies like Alibaba succeed in the enterprise market with open-source large language models, the cards in China’s semiconductor market are likely to be reshuffled in the medium term.