A specific statement made by Jensen Huang at the beginning of 2026 has been cited in financial newsrooms ever since. He stated that Nvidia’s market share in China has completely vanished. not rejected. not crumbled. Nothing.
The statement, which came from the CEO of the most valuable semiconductor company in history, carried the kind of weight that is uncommon in earnings calls. About 95% of the Chinese market for AI accelerators was dominated by Nvidia two years prior. One of the more significant reversals in the contemporary history of technology trade has occurred at that time.
| Topic Snapshot | Details |
|---|---|
| Subject | Nvidia market share collapse in China’s AI accelerator sector |
| 2024 Market Share | Approximately 95% |
| 2026 Market Share | Effectively 0% per Jensen Huang’s own remarks |
| Estimated Annual Revenue Loss | Roughly $10 billion to $15 billion |
| Previous China Contribution | 20% to 25% of data center revenue |
| Restricted Models | H20, H800, A800, alongside flagship H100 and A100 |
| Lead Domestic Competitor | Huawei Ascend chip family |
| Other Rising Players | Cambricon, MetaX, Biren Technology |
| Key U.S. Body | Bureau of Industry and Security |
| Strategic Outcome | Accelerated Chinese AI hardware self-sufficiency |
| Wall Street Stance | Strong BUY rating despite China loss |
There are three parts to the trajectory’s story. The H100 and A100 limitations first appeared in 2022 and 2023. Then came the compromised H800 and A800 models, created especially to meet U.S. export restrictions, which Washington likewise outlawed soon after they were made available.
Lastly, the H20, a further-trimmed model designed for Chinese consumers, was subject to more stringent licensing regulations by the middle of 2025. Every additional limitation was intended to impede China’s advancement in AI. Instead, each one quickened the rate at which Chinese consumers completely ceased purchasing Nvidia.
The financial harm is substantial. In the past, the Chinese market accounted for 20% to 25% of Nvidia’s data center sales, which is what propels the company’s outrageous price. The yearly revenue loss is currently estimated to be between $10 billion and $15 billion.
Even for a company the size of Nvidia, that amount is significant, and investors are starting to consider it as a long-term absence rather than a short-term obstacle. In recent projections, Nvidia has started to regard Chinese sales as zero, treating any future revenue as a windfall rather than a baseline.
The clear winner has been Huawei. For Chinese AI labs and hyperscalers, the Ascend chip line—especially the 910B and the more recent 910C—has evolved from a politely tolerated substitute to a preferred option. Even in China five years ago, there would have been skepticism about Huawei’s ability to seriously compete with Nvidia at the top of AI training. There was no hardware.
There was no software ecosystem. There was a real performance disparity. Despite the difficulties associated with switching from CUDA-based pipelines, that gap has now closed enough for significant Chinese AI companies to dedicate training runs on Ascend silicon.
The historical parallel to the smartphone era is difficult to ignore. Chinese gadget manufacturers and American chipmakers used to be closely related. Then followed the forced shift, the U.S. sanctions, Huawei’s Hisilicon business, and ultimately a local Chinese semiconductor ecosystem that, although still lagging in some areas, is no longer as reliant on American suppliers as it once was. The plot of the AI chip is taking a similar, but quicker, turn. Biren, MetaX, and Cambricon are all developing faster than their initial schedules indicated.
It is more difficult to evaluate the geopolitical calculus. From Washington’s point of view, the export restrictions have accomplished their intended goals. Chinese frontier models are no longer powered by Nvidia’s most sophisticated hardware. Longer term, however, the approach has resulted in a side consequence that some U.S. policymakers quietly acknowledge was predictable.
When cut off, China creates its own. The build speed increases with the severity of the cut. The question of whether the ensuing Chinese chip industry stays a regional player or eventually competes globally, including in regions like Southeast Asia and the Middle East, is still up for debate.
The immediate route is obvious for Nvidia. Expand into the Middle Eastern, European, and American markets, where demand is still almost unquenchable. Develop closer ties with independent AI projects in India, Saudi Arabia, and the United Arab Emirates. For the time being, treat China as a closed door.
Wall Street’s high BUY ratings indicate that the market thinks Nvidia can withstand the setback. The next earnings cycle won’t fully show if that confidence is warranted over the following ten years or if a silent deterioration is just getting started. But the zero is real. Additionally, zeros are difficult to reverse in geopolitics.
