As semiconductors have become a geopolitical hot potato over the past few years, it’s no surprise that Nvidia , the leader in artificial intelligence-related chips, has been caught up in escalating tensions between the U.S. and China. When Nvidia returned to the China market last week , seemingly with the blessing of Washington, it sparked debate over the strategic implications for the U.S.’ dominance in AI and China’s own focus on boosting its domestic chip and tech industry. For the U.S., Nvidia’s return could help cement American strength in AI globally, experts told CNBC. For China, it could buy the country time as it continues on its own path to build Nvidia rivals and keep pace with AI software development. It’s a tricky relationship, underscoring the strategic importance of the graphics processing units (GPUs) that Nvidia designs and that currently underpin the world’s AI. “The relationship is symbiotic but I do believe China needs the U.S. technology more at this moment in time,” Daniel Newman, CEO of Futurum, told CNBC’s “The China Connection” last week. Nvidia’s warnings Earlier this year, the U.S. government restricted exports of Nvidia’s H20 chip to China. The product, a less-advanced version of Nvidia’s leading hardware, had been created to comply with previous U.S. export restrictions. Washington has expressed concerned that these chips could be used in areas such as advancing the Chinese military and China’s own AI industry. Nvidia took a $4.5 billion writedown on the unsold inventory and warned the China restrictions could impact billions of dollars of potential sales. Jensen Huang, the CEO of Nvidia, has been critical of export curbs . He has said it would be a “tremendous loss” for Nvidia not to participate in China and that rival Huawei would be able to fulfil the needs of the country in its absence. Huang has argued that the restrictions could boost China’s domestic semiconductor industry and that it risks eroding America’s technological edge. That message appears to have got through to the White House. Nvidia said last week that it received backing from the U.S. government to resume sales of H20 in China. How the move will benefit the U.S. From a business perspective, Nvidia is expected to gain. But for Washington, the move is more strategic. “We want to keep having the Chinese use the American technology stack, because they still rely upon it,” Commerce Secretary Howard Lutnick told CNBC last week. Nvidia has managed to gain a formidable position in the market for semiconductors required to train and run AI applications, not just because of its hardware but also because of the popularity of its software platform known as CUDA, that developers build on. This creates an “ecosystem” around Nvidia’s products which has proven sticky for its users. The return of the H20 to the world’s second-largest economy will “buy China time” to boost its domestic industry, according to Pranay Kotasthane, deputy director at the Takshashila Institution. “But it will also buy the U.S. companies some respite. China is Nvidia’s largest market and is home to 50 per cent of AI developers according to Jensen Huang. If that path is completely closed, American firms like Nvidia will find it difficult to raise revenues and re-invest them in the next round of research and development,” Kotasthane said. “It might be justifiable to restrain access to the most advanced chips but to expand the scope of the restrictions doesn’t make strategic sense.” China domestic chips in focus Huawei has been China’s leading player in developing AI-focused chips. The country’s technology companies are using some of Huawei’s hardware but the firm has yet to overtake the dominance of Nvidia’s latest chips. One possible outcome is that U.S. export curbs will accelerate China’s domestic efforts. China has been looking to boost its domestic semiconductor industry with a particular focus on AI chips. There are a whole host of startups working on new products in the country. Nvidia’s return to China could slowdown that progress. “If Nvidia’s chips are made available to Chinese firms, it could weaken momentum behind domestic chip projects, cut off capital, and delay progress in domestic Chinese hardware. This retains U.S. tech influence over global AI rails,” Tejas Dessai, director of research at Global X ETFs, told CNBC by email. Ultimately, it all goes back to Nvidia’s software which keeps developers locked into its hardware. “Chinese model developers still prefer to use Nvidia hardware, because the domestic alternative AI stack, particularly the software development environment from Huawei is still difficult to use and lacks the depth and flexibility of Nvidia’s offering,” Paul Triolo, a partner at DGA-Albright Stonebridge Group, told CNBC by email. Can China catch up to Nvidia? Still, China’s direction of travel and its quest for domestic providers of AI chips is unlikely to change. “Eventually Chinese AI model developers will have to transitions to a domestic AI stack,” Triolo said. Nvidia’s chips have proved very effective at training huge AI models that require massive amounts of data to be processed. The actual running of those AI model in products like chatbots is known as inferencing. This process may require a different type of chip, which Chinese tech giants as well as startups are working on. “In chips, China’s opportunity could come when the focus shifts to inference. That’s when demand for lower cost, efficient processors could scale, and we believe custom chip programs from big Chinese tech companies could ultimately serve that demand,” Dessai of Global X ETFs said.
