Nvidia spent months trying to get permission to sell its powerful AI chips to China. Washington finally said yes, the company started shipping them, and then Nvidia suddenly stopped.
So what went wrong?
During its November 17 earnings call, Nvidia confirmed that it had stopped shipping licensed Hopper-generation AI chips to China. The products include the H200, one of the processors Chinese tech companies have been trying to get their hands on.
The announcement was surprising because Nvidia had just earned $2.4 billion from licensed Hopper chip sales to China during the previous quarter.
Interim CEO Colette Kress didn't explain why shipments stopped or whether the company plans to resume them.
But there's an interesting detail here. While the U.S. had approved limited exports, China was restricting how many H200 chips could enter the country.
And if those restrictions continue, Nvidia could end up losing customers to the very Chinese chipmakers it's trying to compete against.
Nvidia got U.S. approval. So why couldn't it keep selling?
Nvidia has been dealing with U.S. export restrictions since 2022, when Washington began limiting China's access to advanced AI processors over national security concerns.
The rules prevented Nvidia from selling some of its most powerful chips to Chinese customers, even as demand for AI computing continued to grow.
Then things started looking up for Nvidia.
In December 2025, President Donald Trump announced that the company could sell H200 chips to approved Chinese customers.
In January, the Commerce Department followed up with a policy that allowed H200 export applications to be reviewed individually.
Of course, Nvidia couldn't simply start shipping chips wherever it wanted.
The company needed to demonstrate that exports wouldn't reduce chip production capacity available to American customers. Chinese buyers also had to meet compliance requirements, and the processors needed security testing in the U.S.
By February, Nvidia had received licenses to ship limited quantities to specific Chinese customers.
After all those negotiations, you'd expect the company to start filling orders.
Except China had other plans.
In an August SEC filing, Nvidia acknowledged that Chinese government restrictions had prevented it from selling all the H200 chips covered by its American licenses.
Essentially, Washington had permitted Nvidia to export chips, but Chinese authorities weren't allowing every shipment into the country.
Around 600,000 Nvidia chips were waiting in Hong Kong
Earlier this month, we got a clearer picture of how complicated the situation had become.
On November 5, Reuters reported that approximately 600,000 H200 chips were sitting in Hong Kong because China hadn't approved their import into the mainland.
That's a huge amount of AI computing hardware waiting for permission to reach customers.
And it wasn't because Chinese companies had suddenly lost interest.
According to Reuters, several major Chinese internet companies received roughly 10,000 H200 chips each in recent months. Smaller shipments were getting through, but large-scale imports remained restricted.
So why would China limit access to processors its technology companies clearly wanted?
One possible explanation is that Beijing wants Chinese AI developers to rely less on American technology.
Think about the position those companies are in. They need powerful chips to train AI models, but Washington can change the rules governing access to Nvidia's products.
And the U.S. has already done so several times.
Chinese-made processors could offer more predictable access, even if companies have to adjust their software or accept differences in performance.
China has encouraged its domestic semiconductor industry for years, and restrictions on Nvidia could give local manufacturers another opportunity to attract customers.
Still, there's something we need to be careful about here.
Nvidia hasn't confirmed that China's import restrictions caused the latest shipment halt. The timing makes the restrictions relevant, but we don't know whether they were the deciding factor.
Nvidia is losing sales, but how much does it hurt?
There's another complication in Nvidia's arrangement with Washington: the chips face a 25% tariff.
Nvidia's H200 processors are manufactured by Taiwan's TSMC. Under the U.S. export arrangement, eligible chips must pass through the United States for inspection before reaching Chinese buyers.
That process exposes them to a 25% tariff, which Nvidia said it hadn't been able to pass on to customers.
So even when the company receives permission to export the processors, it faces additional costs.
But how much money is Nvidia actually losing from the shipment halt?
For context, Nvidia's latest earnings report showed $115.8 billion in quarterly revenue. Its Data Center business alone brought in $107.6 billion.
The $2.4 billion from licensed Hopper products sold to China accounted for approximately 2.2% of Data Center revenue.
That's substantial business, but Nvidia is earning far more from customers elsewhere.
The company also expects $130 billion in revenue next quarter, and its forecast already excludes additional Data Center computing sales to China.
Financially, Nvidia appears able to absorb the immediate loss of those shipments.
The bigger concern is what happens if Chinese companies stop relying on Nvidia altogether.
Could China end up benefiting from Nvidia's restrictions?
Here's the problem Nvidia may face over the next few years.
Chinese technology companies still need processors for their AI models. If they can't buy Nvidia chips, they'll have to look elsewhere.
And domestic manufacturers would love to win those customers.
It's not as simple as swapping one processor for another, though.
Companies may need to modify their software and development processes to work with different hardware. Nvidia also has an established software ecosystem that developers use.
But once a business invests time and money in another chip platform, it may have less reason to return to Nvidia.
Nvidia itself has acknowledged this possibility.
In its August filing, the company warned that export restrictions could help competitors attract customers and expand their software ecosystems.
Chinese chipmakers still face manufacturing and performance challenges, so there's no guarantee they'll replace Nvidia anytime soon.
However, continued restrictions could give them more customers and additional opportunities to improve their products.
For Nvidia, losing a few billion dollars in quarterly sales is one concern. Losing a major customer market over time could be considerably more expensive.
So, why did Nvidia really stop shipping chips?
For now, Nvidia hasn't given us a definitive explanation.
The announcement came shortly after Jensen Huang resigned as CEO, a leadership change The Neuron recently covered. Nvidia has also faced national security scrutiny and established a board committee responsible for export compliance.
But the company hasn't said any of those developments prompted the decision.
China's restrictions are another possible factor, particularly with hundreds of thousands of H200 chips reportedly waiting for import approval.
The U.S. tariff requirements also make those shipments more expensive. We just don't know which factors, if any, led Nvidia to stop shipping.
What we do know is that Nvidia spent months securing permission to sell its processors, generated $2.4 billion from licensed Hopper sales in one quarter, and now expects no additional shipments of those products in the next.
To resume sales, Nvidia needs U.S. export authorization and Chinese import approval. Both governments have a say in whether those transactions can happen.
Meanwhile, Chinese tech companies still need computing power, and domestic chipmakers have an opportunity to win more of their business.
Nvidia can afford to lose the immediate sales. The harder question is whether Chinese customers will still want its processors by the time shipments become easier again.