Nvidia Halts Licensed H200 Chip Shipments to China

· Cortex Solutions

A shipping container of AI server chips held behind a closed barrier at a port at dusk

Nvidia says it has stopped shipping US-licensed Hopper H200 chips to China after $2.4B in Q3 sales, as Congress probes approvals. What it means for Saudi firms.

Nvidia, the world's largest supplier of AI chips, has announced that it has stopped shipping US-licensed "Hopper" chips to China, including the H200 chips that Washington had allowed it to sell to a number of Chinese companies under conditions. The announcement came during the company's third-quarter earnings call on Tuesday, November 17, 2026, and coverage in the following days added details from the US Congress. Here we summarize what the sources reported and what it means for the computing market and for businesses in Saudi Arabia.

What exactly did Nvidia say?

According to a Reuters report on November 18, 2026 (by Eduardo Baptista in Shenzhen), Colette Kress, who took over the company after Jensen Huang's resignation last week, said Nvidia sold $2.4 billion of licensed Hopper-generation data center products to China in the quarter ended October 25. She added that the company's guidance for next quarter includes no licensed Hopper products, that Nvidia is no longer shipping these products to China, and that there are no current plans for further shipments.

AFP reported the same remarks on the same day, describing Kress as "interim CEO", and noted that she gave no reason for the decision and did not say whether the halt is permanent. Pepperstone research strategist Dilin Wu commented that "the timing says quite a lot."

Why did shipments stop?

Nvidia gave no official explanation, but the sources describe several factors around the decision:

  • According to Reuters, the US eased its H200 licensing policy in January, reviewing each request individually under security conditions, and approved sales to 10 Chinese companies including Alibaba, Tencent and ByteDance.
  • Reuters says Beijing has since restricted H200 imports as it seeks to reduce its dependence on US technology, and that around 600,000 H200 chips were stored in Hong Kong awaiting shipment after Beijing withheld approval for large-scale imports.
  • AFP quoted Hong Kong solicitor Chandy Ye as saying the practical impact of the halt is limited, because Chinese authorities have been steering the domestic sector away from H200s since January 2026, and that the new US interim rule on remote access to advanced computing is "the far more consequential development."
  • Washington still bans exports of the newer Blackwell and Rubin GPUs to China, according to AFP.

Congress steps in

On November 19, 2026, CNBC reported that the House Foreign Affairs Committee released documents showing that the Commerce Department's Bureau of Industry and Security had approved, as of May 2026, sales of more than 879,000 AI chips to Chinese companies, out of 3.4 million requests made between December and May. Nvidia spokesperson John Rizzo told CNBC: "Despite receiving some licenses, we have since stopped shipping H200s to China."

CNBC also reported that customers in China and Hong Kong made up about 9.5% of Nvidia's total revenue last quarter, and that Rep. Brian Mast accused two firms named in license applications of being smuggling fronts. It added that the Commerce Department issued an interim rule on Monday to close a loophole on renting Nvidia compute from data centers in third countries, and that closed-door hearings are expected in early December.

The bigger picture

This is not just about one chip shipment; it reflects a new equation in the computing market. Advanced chips have become a political and security file managed directly by governments, and licenses can be granted and then stall because of a decision by another party. The new US rules on remote compute rental also mean that oversight is no longer limited to where a chip sits, but extends to who uses its capacity and from where.

What this means for Saudi businesses

This is Cortex's own reading of what the news means for companies in the Kingdom, not something stated in the sources:

  • AI computing capacity has become a strategic resource subject to international politics, so where your systems and models are hosted deserves a serious discussion at management level, not a passing technical decision.
  • National investment in data centers and cloud infrastructure inside the Kingdom becomes more important in this landscape, because it brings computing closer to local businesses and reduces exposure to supply swings.
  • When building AI solutions for your business, design them so they are not tied to a single provider or chip type; choose an architecture that lets you switch the model or where it runs without rebuilding the whole system.
  • Review your cloud and AI provider contracts for export-control compliance and access, especially if your data or users are in more than one country.

At Cortex we build AI and automation solutions for Saudi businesses with these considerations in mind from day one. Learn more about our services: https://cortex.sa/services/ai-automation

Sources: Reuters (November 18, 2026), AFP (November 18, 2026), CNBC (November 19, 2026).

Read next