NVIDIA share in China drops to 0%
- Jensen Huang said Nvidia is now “100% out of China” and went from roughly 95% market share to 0% after successive U.S. AI-chip restrictions. - The clearest counterpoint is Huawei: its Ascend 950PR entered mass production in March, and existing 2026 orders imply about $12 billion in chip revenue. - The bigger shift is structural — China’s AI stack is localizing faster, so even looser rules may not fully restore Nvidia’s old position.
Nvidia’s China problem is no longer a temporary sales hiccup. It’s turning into a market-structure story. Jensen Huang has now said Nvidia is effectively out of China’s AI-chip market, and that matters because China used to be one of the company’s biggest strategic arenas. The gap is simple — Washington kept tightening export controls, Nvidia kept designing weaker China-specific chips, and Chinese buyers kept moving to domestic hardware instead. Now the replacement cycle looks real. (letsdatascience.com) ### Did Nvidia really say zero? Basically, yes. At a Citadel Securities event, Huang said Nvidia went from about 95% market share in China to 0%, and that the company is “100% out of China” for forecasting purposes. That sounds more extreme than earlier analyst estimates, which had Nvidia still holding meaningful share in 2024 and project(letsdatascience.com)t-share datapoint, but as a blunt statement that Nvidia’s current China AI business has been gutted by policy. (finblog.com) ### What broke the old model? Export controls did. Nvidia’s most advanced AI GPUs were blocked first, then even its watered-down China products faced tighter limits. The H20 had been the company’s main legal workaround for China, but newer restrictions cut into that path too. Huang has been unusually direct here — he’s called the controls a failure and(finblog.com) to slow China also created a giant incentive to replace Nvidia. (cnbc.com) ### Who is filling the gap? Huawei is the obvious winner. Reports this week say Huawei expects AI-chip revenue to reach about $12 billion in 2026, up from $7.5 billion in 2025, based on orders already in hand. The bulk of demand is tied to the Ascend 950PR, which entered mass production in March, with an upgraded 950DT planned for the fourt(cnbc.com)ut it does mean Chinese buyers are no longer waiting around for perfect substitutes. They’re buying what they can deploy now. (finance.yahoo.com) ### Why does software matter so much? Because GPUs are not just chips. They’re ecosystems. Nvidia’s real moat has always been CUDA — the software stack developers already know. That used to make switching painful. But if access to Nvidia hardware keeps shrinking, the pain flips sides. Chinese cloud companies and model builders start(finance.yahoo.com) stops being automatic. This is the part that can outlast the sanctions themselves. (letsdatascience.com) ### Is the “0%” claim the whole story? Not exactly. Earlier forecasts were less dramatic. Bernstein had Nvidia at 66% of China’s AI-chip market in 2024 and projected 54% in 2025, not zero. So there are really two frames here. One is the analyst view — erosion, but not instant disappearance. The other is Huang’s operational view — for the (letsdatascience.com)t in the same direction. Fast decline. (ainvest.com) ### Why does this matter outside China? Because China is too big to lose cleanly. It is a huge pool of AI demand, engineering talent, and future standards-setting power. If domestic Chinese chips get “good enough” at scale, Nvidia does not just lose near-term revenue. It helps create a parallel AI hardware ecosystem that is less dependent on U.S. (ainvest.com)tor market. (cnbc.com) ### So what’s the bottom line? The headline is dramatic, but the real story is slower and bigger. Nvidia did not just lose some sales in China — it may have helped lose the habit of buying Nvidia in China. And once customers rebuild around local hardware, that habit is hard to win back. (finblog.com)
