The Chinese government is reportedly urging domestic firms to move away from using artificial intelligence (AI) chips produced by U.S.-based Nvidia, a pivotal player in the global AI market. According to sources cited by Bloomberg, Chinese regulators have been subtly discouraging companies from purchasing Nvidia’s H20 chips, which were specifically designed to meet U.S. export restrictions.
While not an official ban, China’s regulatory landscape makes such government “guidance” tantamount to policy. The H20, a less powerful alternative to Nvidia’s top-tier AI chips, has been a key product for Chinese firms in AI development. However, Beijing’s push for technological self-reliance is driving firms to prioritise homegrown solutions from companies such as Huawei and Cambricon.
Soft Ban, Big Impact
China has not issued a formal decree prohibiting the purchase of Nvidia chips, but in the country’s highly controlled business environment, even non-binding recommendations often hold considerable weight. The move reflects Beijing’s cautious approach to reducing reliance on U.S. technology without escalating tensions with Washington.
Nvidia, which had already been grappling with U.S. government restrictions on its most advanced AI chips like the A100 and H100, now faces further challenges in maintaining its market share in China. The export bans, introduced by the U.S. in 2022, have been aimed at curbing Beijing’s access to cutting-edge AI technology. In response, Nvidia tailored the H20 chip to comply with U.S. export rules, but China’s latest push could further erode its presence in the lucrative Chinese market.
Domestic Alternatives on the Rise
In its quest for self-sufficiency, China has been pouring vast resources into its semiconductor industry, with a goal of reaching 70% domestic production by 2025. Earlier this year, Beijing announced a new semiconductor investment fund worth 344 billion yuan (£40 billion), aimed at accelerating the development of advanced chips crucial for AI applications.
The pressure to adopt domestic alternatives is already paying off. Chinese tech giant Huawei, in collaboration with its chipmaking subsidiary HiSilicon, has been ramping up production of its Kirin and Ascend chip series. These chips, specifically designed to compete with Nvidia’s offerings, are now being supplied to Chinese AI heavyweights like Baidu. In a significant leap forward, Huawei’s Ascend 910C chip is set to rival Nvidia’s H100, with plans for mass production beginning in October.
Meanwhile, ChangXin Memory Technologies, a leading Chinese memory chipmaker, has also made strides by developing high-bandwidth memory (HBM) chips, another crucial component for AI systems.
U.S.-China Tensions Loom Large
Nvidia’s diminishing sales in China underscore the broader geopolitical tensions that have increasingly spilled into the tech sector. In the second quarter of fiscal 2025, Nvidia’s revenue from China accounted for just 12.2% of its total income, down from nearly a quarter (24.6%) in early 2022. The H20 chip, though compliant with U.S. export controls, may not be enough to offset the loss of market share as Chinese companies pivot towards homegrown solutions.
Samsung Electronics could also feel the ripple effect, as Nvidia’s H20 chips use Samsung’s high-bandwidth memory, a crucial element for AI chip performance.
Strategic Shift for the Future
China’s latest move to foster self-reliance in the AI sector is part of a broader, long-term strategy. The country aims to shield its critical industries from potential future disruptions amid intensifying U.S. efforts to limit access to key technologies. By promoting domestic innovation and supporting its tech companies through substantial subsidies and state guidance, Beijing is signalling its intention to compete head-to-head with the U.S. in the race for AI dominance.
As China’s domestic chip production continues to mature, the global semiconductor landscape could see a profound shift, with U.S. firms like Nvidia losing ground to rising Chinese competitors.
