GOW

2.26010.32%

US chip rule takes effect as Nvidia halts H200 sales to China, $2.4B last quarter

LongbridgeAII'm LongbridgeAI, I can summarize articles.

The US Commerce Department's new limits on advanced AI chips take effect on Thursday, November 19, 2026, the day the rule is published in the Federal Register. Nvidia has already gone further than the rule asks. The company has stopped selling its H200 chips to China, interim chief executive Colette Kress said on its earnings call on Tuesday, according to TechCrunch. Semafor also reported the halt.

Nvidia's own filing shows what that business was worth. Shipments of that generation of data center chip to China were $2.4 billion last quarter, its results commentary says. That is 2.1% of revenue, and it equals about a fifth of what Nvidia books from customers based in China. The forecast Nvidia gave on Tuesday already assumed none of it, which changes how that forecast reads. Take the disclosed China shipments out of last quarter and the outlook implies revenue per week rising 6.4%, where the headline comparison gives 4.2%, TECHi's calculation shows.

What takes effect on Thursday?

The rule is called Addressing Remote Access to Advanced Semiconductors. The Commerce Department filed it on Monday, November 16, and it runs to 117 pages. It takes effect on publication.

It does two things. It requires a US license to export, re-export or transfer advanced AI chips to five countries: Singapore, Malaysia, Indonesia, Thailand and Brazil. Washington's stated concern is that chips sent there can be passed on to China. TECHi reported on Tuesday that customers based in those countries account for 1.3% of Nvidia's revenue at most.

The second part reaches further. The rule also covers data centers in any country that provide large amounts of computing power to companies headquartered in China and in other countries under US export controls. Data centers already serving such customers have 90 days to obtain a license. Counted from Thursday, that deadline is February 17, 2027.

China's commerce ministry has said it will take countermeasures if the rule goes ahead, as TECHi reported on Wednesday.

What did Nvidia stop?

The H200 is a data center chip from Nvidia's Hopper generation, two generations behind the Rubin systems it has just begun shipping. The White House approved its export in January, Semafor reported, and TECHi reported in August that the first deliveries were running far below what the licenses allowed.

Kress told analysts on Tuesday that Nvidia would no longer sell H200 chips to China, according to TechCrunch's account of the call, as relayed by PCMag. Semafor's account says Nvidia halted the sales as it prepared for closer scrutiny from lawmakers. Former chief executive Jensen Huang is expected to testify to Congress next month, Semafor reported. TECHi has not seen a transcript of the call.

Nvidia's written materials do not announce a halt. They contain two sentences on China. One gives the size of the shipments. The other is in the outlook: Nvidia said it is not assuming any data center computing revenue from China.

The decision follows a week in which the company's dealings with China became a governance question. Nvidia's board has set up a committee to oversee export controls, as TECHi reported on Tuesday.

How much revenue is at stake?

Start with the figure Nvidia gives. Shipments of Hopper data center products to China were $2.4 billion in the three months to October 25.

Nvidia's revenue in the quarter was $115.84 billion, so the shipments were 2.1% of the total. They were 2.2% of data center revenue, which was $107.6 billion.

Set against China alone, they matter more. A supplement filed with the results shows $11.05 billion of revenue from customers headquartered in China, including Hong Kong. The Hopper shipments equal 22% of that. Nvidia does not say how the rest is made up.

That China figure had been recovering. It was $2.97 billion in the same quarter a year earlier. Over nine months it is $23.48 billion, against $16.62 billion.

If the $2.4 billion were repeated every quarter, it would be $9.6 billion a year. One comparison gives a sense of scale. In the same quarter, every customer based outside the United States, Taiwan and China bought $1.6 billion from Nvidia between them. The halted China shipments were half as large again as that.

Swipe across to compare all columns. Source: Nvidia results commentary and supplemental financial information for the quarter ended October 25, 2026, filed with the SEC; TECHi calculations. China includes Hong Kong and is measured by customer headquarters; the shipments figure is measured by destination, so the two are not strictly like for like.

What does it do to the forecast?

Nothing, on Nvidia's account, because the forecast never included it. The company expects revenue of $130.0 billion this quarter, plus or minus 3%.

That makes the comparison with last quarter uneven. Last quarter's revenue had $2.4 billion of these China shipments in it. This quarter's forecast has none.

TECHi's report on the results noted that the current quarter has 14 weeks, one more than usual, and that revenue per week would rise 4.2% at the midpoint of the forecast. Remove the disclosed China shipments from last quarter and the base falls to $8.73 billion a week from $8.91 billion. The forecast is $9.29 billion a week. On that basis the increase is 6.4%.

At the bottom of Nvidia's range the increase would be 3.2%, and at the top 9.6%.

So the business outside these China sales is growing about two points faster than the headline arithmetic suggested. The halt removes revenue Nvidia had last quarter, and it does not lower what the company told investors to expect.

The larger risk is the part of the rule that covers data centers. PCMag, citing TechCrunch, reported an estimate from the investment bank Jefferies that the restrictions on remote access could cost Nvidia as much as $62 billion of data center revenue next year. Kress said on the call that demand for the Rubin chips would cover a shortfall, according to the same account. For scale, the growth target Kress gave on Tuesday implies revenue of about $741 billion next fiscal year, on TECHi's calculation. The Jefferies figure is 8% of that.

How are the five countries responding?

Carefully. Indonesia's government communications agency said in a statement to the broadcaster CNA on Wednesday that it is reviewing the measure and asking industry how it would affect operations and investment, including in data centers. It said it is open to technical talks on how the rule will apply.

Singapore's prime minister, Lawrence Wong, said the country is prepared to help the United States check how advanced chips are used there, CNA reported. Malaysia has told data centers operating in the country that they must comply with the US rules, Reuters reported.

None of the three has said it will resist. Their data center industries run on American chips.

What does it mean for Nvidia stock?

On the numbers, the H200 decision is small. It is 2% of revenue and it was already outside the forecast. Nvidia closed at $150.96 on Wednesday, its lowest close since June 2025, as TECHi reported.

The case for concern is that the rule is aimed at something harder to measure. Nvidia does not say how much of its revenue comes from data centers abroad that rent capacity to Chinese companies. Customers based outside the United States, Taiwan and China accounted for $1.6 billion last quarter, 1.4% of revenue, but a data center's customers and its owner's headquarters are not the same thing.

The case against is that Nvidia is selling everything it can make. Kress told analysts that demand still far exceeds supply, CBS News reported, and the forecast excludes China and still rises.

What to watch

The 90-day period for existing data centers ends on February 17, 2027. Nvidia's quarterly report, due by December 4, should say more about China than the results release did. Huang's expected appearance before Congress next month will keep the company's past sales there in view, and China has yet to say what its countermeasures will be. TECHi's Nvidia quote page carries the live price.

Source: TECHi The copyright of this article belongs to the original author/organization.

The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.