What Nvidia said about the H200 China halt
Nvidia has stopped shipping its licensed H200 AI chips to China, and it disclosed the Nvidia H200 China halt on its November 17 earnings call. The move came days after a new US export rule, and it has drawn congressional attention to how many chips Washington approved for Chinese buyers.
Interim CEO Colette Kress made the core statement. "We are no longer shipping licensed Hopper data center products to China," she said, as Gizmodo quoted her. The halt covers the H200 and related Hopper-generation products that Washington had licensed for sale to China, according to TweakTown.
Nvidia also removed that revenue from its forward guidance, so the outlook it gave investors no longer includes China, according to BigGo Finance.
The company's explanation is thin. Nvidia spokesperson John Rizzo told CNBC that "Despite receiving some licenses, we have since stopped shipping H200s to China." That confirms licenses were issued and shipments stopped, but it does not say why. Kress did not explain the reasons on the call, and she did not say whether the halt is permanent, according to International Finance.
Why the timing follows the new overseas data center rule
The disclosure came one day after a rule that changes how Chinese firms can use American chips. Commerce issued the rule on Monday, November 16, local time, the day before Nvidia's earnings call. It restricts remote computing access by Chinese companies to American chips through overseas data centers, according to BigGo Finance and Memeburn.
Memeburn reported that Nvidia said it needed time to study that rule. That is a secondhand account, not a direct quote, but it points to a practical question. If Nvidia is still working out how the rule applies to its existing licenses, the halt could be a pause rather than an exit. No source in the research has confirmed that reading.
The sequence also fits a pattern. The administration has licensed sales while taking a cut, then tightened rules on remote access, and the overseas data center rule is the latest step in that approach. USA 24 has covered a separate threat of a Trump tax on AI data centers overseas, which still has no rate or start date. Both stories raise the same question: how Washington plans to police AI computing that sits outside US borders.
How many H200 chips were approved, and how many reached China
The biggest number in this story is also the least settled. Rep. Mast said Commerce approved sales of more than 879,000 AI chips to Chinese companies, according to CNBC. Approval is not delivery, and the reporting shows that gap clearly.
Reuters reported, in an account carried by International Finance, that about 600,000 H200 processors were held in Hong Kong. They were waiting for approval to ship to mainland China because Beijing had withheld approval for large-scale imports. Per CNBC, the US licensed H200 sales to Alibaba, Tencent, ByteDance and other Chinese companies, but sales stayed limited for the same reason.
The policy path has been bumpy. In December, President Trump said Nvidia could ship H200s to approved Chinese customers if the US received a 25% cut of sales. In January, Washington eased its licensing policy so H200 export applications would be reviewed case by case with security conditions. Beijing then restricted H200 imports as it pushed domestic AI hardware.
Memeburn also described a January 13 Commerce rule that routed H200 shipments through the US so the government could collect a 25% fee. That detail comes from a single report. The research for this story did not confirm whether the fee structure still applies, so treat it as unverified.
What the halt means for Nvidia's China revenue
Nvidia booked about $2.4 billion from Hopper shipments to China in its fiscal third quarter, which ended in late October. The table below gathers the key figures and their sources, because several of them are reported differently from outlet to outlet.
| Figure | What it measures | Reported by |
|---|---|---|
| About $2.4 billion | Hopper shipments to China in the quarter ended late October | BigGo Finance, TweakTown |
| About 2% of quarterly revenue | China share of total revenue, per one report | BigGo Finance |
| 2.1% | China share, base not specified | Techi |
| About 2.2% | China share of data center revenue | Memeburn |
| More than 879,000 | AI chips Commerce reportedly approved for Chinese companies | CNBC |
| About 600,000 | H200 processors reportedly held in Hong Kong | Reuters, via International Finance |
The revenue share is small, which limits the direct hit to Nvidia's top line. The reports disagree on the exact percentage, so treat "about 2%" as a rounded estimate. The bigger issue is what the halt signals. A company can hold US licenses and still stop shipping, which makes an export license a weaker guarantee of sales than it first appears.
Market context matters too. The China change arrives during the broader AI stock selloff that USA 24 has covered, so investors will weigh this guidance change against a wider slide in AI shares.
Why Congress is asking about license approvals
Congressional pressure is building. The 879,000 figure raises a basic question: how many licenses were issued, and how many of those resulted in chips reaching Chinese customers? The reporting reviewed here does not answer either part.
Reporting on Jensen Huang's status is weaker. Gizmodo's headline refers to Huang's departure, and Colette Kress spoke on the call as interim CEO. The only source in this research that gives a November 13 resignation date is a newsletter digest from The Neuron, and no other reporting here confirms it. Until that changes, treat the resignation as unverified.
What business readers with China exposure should check
Consumers have no direct action to take here. Business readers with cloud, chip or data center exposure to China or Southeast Asia should review their export-compliance position under the new rule. This is general information, not legal advice.
A practical starting list:
- Map where your AI compute runs, including overseas providers and any customers with ties to China.
- Ask cloud and data center vendors in writing how they screen remote access by Chinese companies to American chips.
- Check contracts and capacity plans that assume Hopper-class hardware will be available in Asia.
- Read the rule text itself, and bring in export counsel before you change who can reach your systems.
The exact scope of the rule matters more than any news summary, including this one.
What to watch next as Washington and Beijing respond
Three questions will shape what comes next. First, whether the halt is permanent or a pause while Nvidia studies the overseas rule. Second, whether Commerce publishes approval numbers that reconcile with what actually reached China. Third, whether Beijing relaxes its import restrictions, which would shift the story from Washington's controls back to Chinese purchasing decisions.
Nvidia's next quarterly report will show whether any China revenue returns to the outlook. Any congressional follow-up on the 879,000 approvals will test how much of that count became real hardware. The Federal Reserve has already flagged the AI buildout as an inflation and stability risk, which gives the export fight a wider backdrop than one chipmaker.




