
Reuters Finds $3.2 Billion in China AI Projects Built on NVIDIA Chips
Reuters documented $3.2 billion in Chinese AI projects using restricted NVIDIA H100, H200 and H800 chips, four quarters after NVIDIA zeroed China out of its guidance.
- By
- Sophie Lindqvist
- Filed
- Channel
- Chips & Policy
- Read
- 4 min read
Chinese organizations have committed more than 23 billion yuan ($3.2 billion) to AI computing projects built around restricted NVIDIA processors, according to a Reuters investigation published November 13, 2026. That is the same NVIDIA whose chief executive, Jensen Huang, declared in 2025 that his company's China market share had collapsed from 95% to zero under U.S. export controls.
The two claims are not necessarily contradictory. But the gap between them now sits at the center of the NVIDIA story, in the same week Huang resigned as CEO with no permanent successor named.
What did NVIDIA tell investors about China?
For four straight quarters, NVIDIA's guidance excluded China entirely. The filings were blunt. In Q3 FY2026, the company said H20 sales were "insignificant in Q3, reflecting export restriction impacts on China-targeted products."
In Q1 FY2027, NVIDIA shipped no H20 compute products to China, against $4.6 billion a year earlier. On the August 26 earnings call, CFO Colette Kress said: "Given ongoing geopolitical uncertainty, there is no China data center compute revenue in our forward outlook." She added that licensed Hopper 200 shipments to China made up less than 1% of Data Center revenue and were "dilutive to corporate gross margins." NVIDIA's August filing stated the company was effectively excluded from China's data-center computing market.
That is the legal picture. Reuters found a different commercial one on the ground.
What did the Reuters investigation find?
Reuters reviewed more than 150 Chinese procurement notices, investment plans, contracts, and corporate reports. Together they document over 23 billion yuan in AI computing projects built around restricted NVIDIA processors, including the H100, H200 and H800. Some projects are running. Some are under construction. Some remain on paper only.
One caveat matters here. The $3.2 billion figure includes supporting infrastructure, so it measures total project spending by Chinese buyers, not NVIDIA chip revenue. It is a broader number than what NVIDIA would report.
Reuters also cited IDC estimates that NVIDIA held 34% of Chinese AI-accelerator shipments in the first half of 2026. That figure counts hardware entering China, including units NVIDIA may not have sold directly or legally. Reuters has not linked these projects to the separate Megaspeed claims.
Read together, Huang's "zero" described NVIDIA's lawful access to China. Chinese demand for its silicon never stopped.
Why did losing China barely dent the numbers?
Huang himself supplied the explanation. Speaking about fiscal 2028, he said: "Even though our demand is much greater than 70%, our supply allows us to confidently deliver 70%." He added: "The unconstrained number would be a lot higher."
The results bear that out. Q2 FY2027 revenue reached $96.22 billion, up 105.8% year over year and ahead of the $92.07 billion consensus. Non-GAAP EPS of $2.22 beat the $2.09 estimate. Data Center revenue was $89.02 billion. Q3 guidance of $108.0 billion, plus or minus 2%, assumes no China Data Center compute revenue at all. "Now, compute is revenue," Huang said.
In other words, NVIDIA could absorb the loss of China because demand elsewhere exceeded what its supply chain could ship. Zero China revenue did not stall growth.
Why does this matter for the stock now?
NVIDIA closed at $162.85 on November 13. That was down 12.75% for the week and 26.41% for the month, a decline that coincided with the leadership change and the Reuters findings.
The investor logic cuts both ways. Because guidance already assumes nothing from China, any licensed reopening would be pure upside to the forecast. But the same reporting shows restricted NVIDIA chips running inside active Chinese projects, and Reuters reports that the leadership transition and U.S. scrutiny now imperil those projects. That concentrates export-control risk in a company that has already told investors to expect nothing from China.
What is not said in NVIDIA's filings is the question Reuters leaves open: how much of the $3.2 billion in Chinese spending flowed back, directly or indirectly, into hardware the company says it cannot lawfully sell. The commercial picture now depends on Washington's next move, and on whether Huang's successor keeps China at zero in the guidance model.
Source: 247wallst.com
More from Sophie Lindqvist
Show full bio
News editor covering money, funding and market shifts around Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,004 articles
Related articles
alleged-nvidia-7-5-billion-china-export-control-gap
Alleged Nvidia $7.5 Billion China Export-Control Gap
nvidia-reaffirms-260-000-gpu-supply-to-south-korea-after-huang-exit
Nvidia Reaffirms 260,000-GPU Supply to South Korea After Huang Exit
china-reportedly-mulls-allowing-nvidia-ai-chip-sales-to-bytedance-alibaba-4fd2374b
China Reportedly Mulls Allowing Nvidia AI Chip Sales to ByteDance, Alibaba
jensen-huang-exit-threatens-3-2-billion-in-china-ai-projects
Jensen Huang exit threatens $3.2 billion in China AI projects


