Nvidia has stopped shipping licensed Hopper data-center products to China, interim CEO Colette Kress disclosed during the company's November 17, 2026, earnings call, ending a sales channel that had generated $2.4 billion in a single quarter despite holding U.S. export approval.
Nvidia's Q4 fiscal 2027 outlook targets $130 billion in revenue but explicitly assumes no China data-center compute contribution. TechCrunch identifies the affected hardware as H200s. Separately, new U.S. rules reportedly bar advanced chips from foreign data centers serving Chinese customers remotely, with Jefferies estimating a potential $44–$62 billion impact on Nvidia's data-center revenue next year.
Nvidia sold $2.4 billion of licensed Hopper data-center products to China in its third quarter, but those shipments have now stopped. Interim CEO Colette Kress disclosed the change during the company’s November 17, 2026, earnings call, according to Bloomberg Law’s reporting.
The company’s official forecast makes the financial shift explicit. Nvidia expects $130 billion in revenue, plus or minus 3%, for its fourth quarter of fiscal 2027, with no contribution assumed from China data-center compute sales. Its November 17 earnings release states:
“NVIDIA is not assuming any Data Center compute revenue from China in its outlook.”
The reversal removes an operating sales channel that had received U.S. approval. For Chinese AI labs and infrastructure buyers, the immediate issue is supply: an export license alone no longer means Nvidia is shipping the hardware. For domestic chip suppliers, that creates an opportunity to win purchases and software investment that might otherwise have gone to Nvidia.
What Nvidia Has Actually Stopped
Kress’s earnings-call statement describes an operational halt, alongside the more cautious language of a financial forecast. According to Bloomberg Law, she said:
“we are no longer shipping licensed Hopper data center products to China”
TechCrunch identifies the affected chips as H200s. The distinction in terminology is useful: H200 is the product name in that report, while Kress’s quoted statement refers to the broader licensed Hopper data-center category.
The forecast has a wider scope still. It excludes any China data-center compute revenue, not only the licensed Hopper products discussed on the call. Taken together, the disclosures show both a stopped shipment channel and a next-quarter plan that does not depend on China compute sales.
Nvidia has not described the halt as a permanent withdrawal from every future China data-center opportunity. Nor is the outlook a statement that all Nvidia revenue connected to China, across every product category, will disappear. The supported conclusion is narrower: licensed Hopper shipments have stopped, and the company is budgeting for no China data-center compute revenue in the coming quarter.
I’d treat the shipping statement as the most important procurement signal. Forecasts can change, but buyers planning a new cluster need deliverable equipment, not simply a product that previously qualified for an export license.
A U.S.-Approved Sales Route Stops Operating
The H200 arrangement had offered Nvidia a way to sell data-center chips into China despite broader restrictions on advanced AI hardware.
According to TechCrunch’s account, former CEO Jensen Huang and President Donald Trump brokered the arrangement. It required the chips to pass through the United States, allowing the government to collect a 25% tariff. Bloomberg Law separately reports that Nvidia had received U.S. approval to sell a version of its Hopper products to China.
That history makes this development different from a straightforward announcement of a new export ban. Nvidia had an approved route and recorded sales through it. The company now says it is no longer shipping those licensed products.
The available disclosures do not establish that the licenses themselves have been revoked, or identify a single definitive reason for the stop. Calling the development a license cancellation would confuse permission to export with the commercial decision and practical ability to deliver.
Chinese acceptance was already a separate obstacle. TechCrunch reports that China had largely refused H200 deliveries even before Nvidia’s halt. It cites Jacob Feldgoise, a researcher at Georgetown’s Center for Security and Emerging Technology, on the government’s preference for directing demand toward domestic chipmakers.
The route therefore faced pressure from both sides: U.S. permission came with conditions, while Chinese industrial policy favored local alternatives.
The Revenue at Stake Is Significant, but Specific
Nvidia’s results show why it can remove China compute revenue from its forecast while still projecting overall growth.
The company reported $115.8 billion in revenue for the third quarter of fiscal 2027, which ended October 25, 2026. Data Center contributed $107.6 billion. Those figures were up 103% and 110%, respectively, from a year earlier, according to the official earnings release.
Measure
Q3 FY2027 reported result
Q4 FY2027 outlook
Total revenue
$115.8 billion
$130 billion, plus or minus 3%
Data Center revenue
$107.6 billion
No separate segment forecast in the release
Licensed Hopper sales to China
$2.4 billion, according to Kress as reported by Bloomberg Law
No licensed Hopper products included, according to Kress
China data-center compute revenue
The licensed Hopper figure is one disclosed component
No revenue assumed
Using the reported figures, the $2.4 billion in licensed Hopper sales equaled approximately 2.1% of Nvidia’s total quarterly revenue and 2.2% of Data Center revenue. Those percentages are calculations from Nvidia’s results and Bloomberg Law’s reported China sales figure, not market-share estimates.
The scale matters in two ways. A multibillion-dollar quarterly sales channel is consequential for suppliers and customers. Yet it is a relatively small part of Nvidia’s current business, which helps explain how the company can forecast higher revenue without it.
At the midpoint, the $130 billion outlook represents approximately 12.3% sequential growth. Nvidia notes that the fourth quarter includes an additional week, so that increase should not be read as a clean measure of growth at an unchanged weekly pace.
The $2.4 billion also needs careful handling. It is revenue from a completed quarter, not a disclosed cancellation total or a guaranteed amount Nvidia would otherwise have earned next quarter. It provides a concrete measure of the channel’s recent size, but it cannot establish the exact future revenue lost.
Domestic Suppliers Gain a Chance to Win More Than Orders
Chinese chipmakers gain a competitive opening because a previously available Nvidia purchase option has disappeared from current shipments. That is a supply-side change, not proof that domestic alternatives have matched Nvidia’s performance.
TechCrunch’s reporting on China’s resistance to deliveries suggests that the government was already trying to move demand in this direction. Nvidia’s halt strengthens that pressure: buyers considering local hardware now face fewer dependable options for purchasing new Nvidia compute through the licensed route.
For frontier-model developers, replacing a planned accelerator purchase involves more than choosing a different chip. Large training jobs distribute computation across many accelerators. Their practical performance depends on memory capacity, communication between devices, software support and how reliably the whole system runs.
Those dependencies create work when a lab changes platforms. Teams may need to adapt software, validate numerical behavior, tune distributed training and revise deployment plans. These are practical implications of switching computing platforms, not findings from tests of any particular Chinese product.
That work can also create a lasting benefit for the alternative supplier. Once a customer has invested in making its models run efficiently on another platform, a future purchase has less migration work attached to it.
My view is that this software investment is the more consequential competitive opportunity. A single hardware order brings revenue; a customer that builds and maintains a usable alternative gains a reason to consider that supplier again.
The evidence does not identify which domestic vendors will capture the displaced demand, how much hardware they can deliver, or whether their systems can satisfy particular frontier-training requirements. Some buyers may delay expansion or change their plans. The opening is real, but the eventual distribution of orders remains unknown.
Overseas Compute Access Faces Separate Pressure
The shipment halt arrives alongside another reported restriction on Chinese access to advanced Nvidia hardware.
TechCrunch reports that U.S. rules issued the previous day bar advanced U.S. chips from foreign data centers serving Chinese customers remotely. That matters because the location of a chip and the location of its customer can differ: a Chinese lab could seek computing capacity from a provider operating outside China.
Export-control attorney Dan Fisher-Owens told TechCrunch that large data centers suited to AI training would have to certify that they were not serving Chinese customers to buy Nvidia chips. In that account, providers unable to make the certification could face pressure to consider other chip options.
This is a separate mechanism from stopping licensed H200 shipments into China. One concerns physical deliveries to the country; the other concerns access through overseas infrastructure. Together, they could narrow buyers’ alternatives more substantially than either change alone.
The financial interpretations differ. TechCrunch reports that investment bank Jefferies estimated the new rules could trim between $44 billion and $62 billion from Nvidia’s data-center revenue next year. That is an analyst forecast about the broader rules, not a booked loss or an estimate specifically for the H200 shipment halt.
Kress, by contrast, said demand for advanced products such as Vera Rubin meant the rules would have no impact on Nvidia’s top-line revenue, according to TechCrunch. The available reporting does not provide enough detail to reconcile the assumptions behind those assessments. The near-term company outlook and the analyst estimate also address different time horizons.
What Chip Buyers Should Take From the Disclosure
Chinese labs now have a concrete reason to separate three questions in their infrastructure plans:
Export permission: Has a transaction received the necessary regulatory approval?
Actual supply: Is the vendor accepting and fulfilling shipments?
Usable capacity: Can the delivered system run the intended workload at the required scale?
Nvidia’s disclosure shows why the first answer cannot substitute for the second. The platform-switching work shows why the second cannot substitute for the third.
For Nvidia, strong demand elsewhere may protect near-term revenue while its position in China changes. Its official release says Vera Rubin NVL72 systems have begun shipping and names cloud partners running the platform. That supports the company’s account of an expanding business outside the stopped Hopper channel, although future demand remains a forecast.
For Chinese suppliers, the opportunity is to turn constrained customer choice into working deployments. Their most durable gains would come from proving they can deliver systems, support software and keep large workloads running. Nvidia’s halt gives them a stronger reason to be evaluated; it does not settle the outcome of that evaluation.
Frequently Asked Questions
3 questions
1
Is Nvidia Permanently Exiting China?
Nvidia has not announced a permanent exit from every China data-center opportunity. It says licensed Hopper shipments have stopped, while its Q4 outlook assumes no China data-center compute revenue.
2
How Much Licensed Hopper Revenue Came From China?
Nvidia sold $2.4 billion of licensed Hopper data-center products to China in Q3 FY2027, according to Kress’s earnings-call comments reported by Bloomberg Law. That was approximately 2.1% of total quarterly revenue.
3
Do the Reported U.S. Rules Also Affect Overseas Access?
Yes. TechCrunch reports that the new rules restrict advanced U.S. chips at foreign data centers serving Chinese customers remotely. That is distinct from the halt in licensed Hopper shipments directly to China.