Why Is Nvidia Courting China Despite US Chip Sanctions?

In a move that Beijing hailed as a “vote of confidence,” US semiconductor giant Nvidia confirmed it will attend China’s flagship supply chain expo in July. The appearance comes even as Washington intensifies a campaign to hobble China’s technological advancement, implementing sweeping export controls aimed squarely at Nvidia’s advanced artificial intelligence chips.

The situation highlights the tightrope Nvidia is walking between a massive commercial opportunity and the geopolitical objectives of its home country. With a formidable domestic rival in Huawei rapidly gaining ground, Nvidia faces a difficult question: how does it compete in a critical market that its own government is trying to cut it off from?

How important is the China market to Nvidia?

China represents a colossal, if complicated, prize. Nvidia CEO Jensen Huang has estimated China's AI market could be worth $50 billion within a few years, calling the prospect of missing out a “tremendous loss.” Before the toughest restrictions took hold, China accounted for as much as 20-25% of Nvidia’s business. The country is also a vital hub for talent, with Huang noting that China has “50% of the world’s AI developers.”

The strategic argument, as voiced by Huang, is that it's crucial for these developers to build on an American technology stack, primarily Nvidia’s CUDA software platform. Ceding this ground entirely allows Chinese competitors to establish their own ecosystems, which could then challenge the US globally. “When China starts to aggressively diffuse their AI technology, so long as all the AI developers are in China, I think China's stack is going to win,” Huang warned in a June interview.

What are the US restrictions and what's their impact?

Starting in 2022 and escalating since, the US Commerce Department has imposed stringent export controls to slow China’s AI progress and military modernization. The rules restrict the sale of high-performance GPUs based on metrics like computing power (flops) and interconnect bandwidth, directly impacting Nvidia’s top-tier chips like the H100.

Nvidia responded by creating a series of downgraded chips specifically for China, such as the H800 and H20, which were designed to fall just below the regulatory thresholds. However, the Trump administration tightened the rules again in April 2025, effectively banning the H20 as well.

The financial blow has been significant. In its May 2025 earnings report, Nvidia announced it took a $5.5 billion charge on its now-restricted H20 inventory and projected an $8 billion hit to its quarterly revenue outlook due to the lost opportunity in China. “The H20 export ban ended our Hopper data center business in China,” Huang stated bluntly.

How is China responding to the sanctions?

The US strategy has had the unintended consequence of catalyzing China's domestic semiconductor industry. Huawei has emerged as Nvidia's most formidable challenger. Cut off from Western technology, Huawei has focused on developing its own AI ecosystem, from its Ascend series of AI processors to its CANN software architecture.

Huawei's progress has been startlingly fast. CEO Jensen Huang himself has acknowledged that Huawei is a “formidable technology company” and that its latest Ascend chips are now comparable to Nvidia's H200. While individual Chinese chips may lag in power efficiency, Huawei is competing at a systems level. Its CloudMatrix 384, a computing system connecting 384 of its Ascend 910C chips, reportedly outperforms Nvidia's GB200 NVL72 rack on some metrics, though it consumes significantly more power.

This hardware is being produced domestically by Semiconductor Manufacturing International Corp. (SMIC), which, despite being blacklisted and cut off from ASML’s most advanced EUV lithography machines, has managed to produce 7-nanometer chips. This progress ensures that major Chinese cloud companies like Alibaba, Tencent, and ByteDance, which were once major Nvidia customers, have a viable domestic alternative to turn to.

Can Nvidia still compete in China?

Despite the restrictions, Nvidia is not giving up. The company is reportedly preparing to launch a new, even more constrained chip for the Chinese market based on its latest Blackwell architecture. The processor is expected to use conventional GDDR7 memory instead of high-bandwidth memory (HBM) and will not require TSMC’s advanced CoWoS packaging, simplifying its manufacturing and ensuring it complies with US rules.

Nvidia’s most durable competitive advantage remains its CUDA software platform. For nearly two decades, CUDA has become the de facto standard for AI development, and millions of developers are trained on it. This entrenched ecosystem is difficult for competitors to replicate. While Huawei’s CANN is improving, it is playing catch-up against a deeply established incumbent.

The challenge for Nvidia is to offer a product that is both compliant with US law and compelling enough for Chinese customers to choose over rapidly improving domestic options. The company's recent expo appearance signals its intent to stay engaged, but its ability to do so effectively remains caught between Washington's restrictions and Beijing's push for self-sufficiency.

Reference Shelf

• China hails Nvidia expo appearance as ‘vote of confidence’ amid tensions (SCMP)
• Is Huawei a Real Threat to Nvidia’s AI Empire? (ARPU)
• Nvidia CEO Jensen Huang on AI, Musk and Trump (Bloomberg Television)
• Nvidia's CEO says China is not far behind the U.S. in AI capabilities | Tom's Hardware (Tom's Hardware)