Nvidia’s China Market Share Nears Zero: Amid Rise of Domestic Computing Power, Does Jensen Huang Still Have a Chance for a Comeback?

Nvidia's setback in the Chinese market is essentially a passive outcome of the combined effects of US-China geopolitical rivalry and export control policies, rather than a decline in its technological competitiveness. The industry giant, which once controlled a 95% share of China's high-end AI chip market, has now been forced to accept the reality that sales of its core products in China have ground to a near-halt.

Since 2022, the Bureau of Industry and Security of the US Department of Commerce has continuously tightened semiconductor export restrictions to China, moving from direct bans on top-tier flagship GPUs like the A100 and H100, to blocking downgraded versions like the A800 and H800 designed to circumvent controls, and recently to implementing stringent 'look-through' whitelist reviews on Asian clients. Consequently, Nvidia's official distribution channels targeting the Chinese market have been completely severed.

For major Chinese cloud service providers, under such a stringent regulatory environment, the compliance risks of purchasing Nvidia chips have far outweighed their commercial value, forcing them to turn to other solutions to ensure the stability of their computing power supply.

To preserve its Chinese market share as much as possible, Nvidia once introduced the H20 chip, customized specifically for China. However, this 'China-specific' product, with its severely stripped-down computing power, did not achieve the expected results.

The AI training performance of the H20 was severely degraded compared to its predecessor, failing to meet the high-performance demands of leading Chinese large language model developers. More crucially, due to the uncertainty of US regulatory policies, Chinese companies are unwilling to bet massive capital expenditures on a product that could face supply cutoffs at any time.

Ultimately, this product also ran into obstacles with regulatory approvals and faced halted shipments, forcing Nvidia to directly assume 'zero China sales' in its revenue guidance. CEO Jensen Huang explicitly stated during an earnings call that the company will plan for the next several quarters under the assumption of zero sales in the Chinese market, effectively conceding the reality that it has temporarily lost its dominant position in China's advanced AI market.

The massive market space left by Nvidia's retreat is being rapidly and fully filled by fast-growing domestic Chinese AI chipmakers.

Domestic AI chips centered on Huawei's Ascend series have gradually approached the performance level of Nvidia's A100 in computing power, making key breakthroughs particularly in cluster networking capabilities.

The Ascend 910B and its subsequent iterative 910C chips not only surpass Nvidia's H20 chip, which was customized for the Chinese market, in single-card computing performance, but their supporting full-stack software ecosystem (such as the CANN operator library) is also continuously improving, substantially reducing the technical and time costs for enterprises migrating from overseas platforms to domestic computing power.

In addition to Huawei, Hygon Information's DCU chips and Cambricon's Siyuan series chips have also shown strong competitiveness in specific scenarios, providing a diverse range of computing options for enterprises with different needs.

Facing the computing power crisis caused by chip shortages, Chinese internet giants have adjusted their strategies, shifting from a single architecture heavily dependent on Nvidia to fully embracing domestic computing power. Leading companies such as Baidu and Tencent have already deployed Huawei Ascend clusters on a large scale for large model training, while others like ByteDance are actively testing and adopting domestic chip solutions.

This demonstration effect from leading clients has completely revitalized the supply chain ecosystem of domestic AI chips, prompting more small and medium-sized enterprises to adopt domestic computing power. According to the latest survey report from Bloomberg Intelligence, Chinese enterprises plan to allocate 46% of their AI accelerator budgets to local products over the next 12 months, far exceeding the current 30%, with domestic manufacturers like Hygon Information and Cambricon becoming new market focal points.

Even as chip sales in China have nearly ground to a halt, Jensen Huang has repeatedly emphasized the strategic value of the Chinese market in various public forums, stating bluntly that China is one of the world's largest AI markets, boasting a massive pool of AI developers, and its AI chip market is expected to expand from approximately $50 billion currently to hundreds of billions of dollars. If U.S. companies are excluded, they will miss out on this market with immense growth potential, which would undoubtedly be a major loss.

To this end, Huang has been communicating with the U.S. government, hoping to secure export licenses for customized GPUs—including China-specific versions based on the Blackwell architecture—while simultaneously accelerating the rollout of AI infrastructure in other regions globally, attempting to temporarily offset the shortfall in China with markets in Europe, the U.S., and the Middle East.

However, whether Nvidia can return to the Chinese market remains constrained by three core variables.

First is the policy level: whether the U.S. will moderately adjust its high-end chip export policies to China in the coming years to leave room for Nvidia to return to China's high-end AI market is a key prerequisite determining the possibility of its return.

Second is the technical level: whether Nvidia can maintain its globally leading architectural advantage while developing customized products that balance performance, power consumption, and compliance requirements—rather than simply rolling out performance-hobbled chips—will directly impact its competitiveness in the Chinese market.

Finally, at the market level, if domestic chips have already formed a stable, closed-loop ecosystem across cloud service providers, large language model developers, and government-enterprise scenarios, Nvidia's market space will be significantly compressed even if it is allowed to return.

Judging from the current landscape, U.S. export policies remain highly uncertain, and the trend of domestic substitution continues to strengthen. The probability of Nvidia achieving a "complete turnaround" in the Chinese market in the short term is extremely low. A more realistic path is to gradually restore its market share in the mid-to-low-end or specific niche segments, making it difficult to replicate its past dominance when it held 95% of the high-end GPU market.

If policies ease marginally in the future, Nvidia is more likely to participate in the Chinese market in a collaborative manner, rather than relying solely on hardware sales.

In all likelihood, Jensen Huang will choose to deepen cooperation with domestic cloud vendors, ICT giants, and ecosystem partners, deeply integrating into China's AI infrastructure construction through diverse models such as joint solutions, hybrid architecture deployments, and open toolchains, thereby transitioning from a traditional hardware supplier to an ecosystem participant.

For Nvidia, losing the Chinese market, which historically accounted for more than 20% of its revenue, means it must rely more heavily on cloud service providers in North America and other regions globally to absorb its capacity, significantly limiting its growth potential.

Meanwhile, for China's AI industry, this supply disruption crisis has instead become the strongest catalyst forcing the domestic semiconductor supply chain to achieve technological leaps.

History has long proven that technological blockades are never a long-term solution—the tighter the restrictions, the more they spur independent innovation. The dramatic shift in Nvidia's Chinese market is not merely a matter of market gains and losses for a single company, but a microcosm of the evolving global technology landscape.

This content was translated using AI and reviewed for clarity. It is for informational purposes only.