Regulation•

Washington Cleared Nearly 900,000 Nvidia and AMD AI Chips for China in Six Months, Out of More Than 3.4 Million Requested

A House committee says US officials approved nearly 900,000 Nvidia and AMD AI chips for Chinese buyers including Tencent and Alibaba in six months. How many shipped is unknown.

Written By :MINRK
Washington Cleared Nearly 900,000 Nvidia and AMD AI Chips for China in Six Months, Out of More Than 3.4 Million Requested

Photo by Michael Dziedzic on Unsplash

1. House panel puts a number on approved chip exports

The US House Foreign Affairs Committee disclosed on November 19, 2026 that the United States approved the sale of nearly 900,000 artificial intelligence chips to China over a six-month period. The figures on Nvidia AI chips for China, along with processors from Advanced Micro Devices (AMD), came from export requests covering more than 3.4 million units, with Tencent and Alibaba among the intended buyers.

The disclosure matters because advanced AI processors sit at the center of the technology rivalry between Washington and Beijing. For several years the United States has restricted the most capable chips from reaching Chinese customers on national security grounds. The committee's numbers show that, inside that system of restrictions, a large volume of sales was nonetheless authorized.

One important question is left open. The committee's figures describe what the government allowed, not what actually crossed the border. It is not clear how many of the approved chips were delivered, in part because Nvidia stopped its shipments even though it held licenses to make them.

2. What the Nvidia AI chips China license data shows

Two headline numbers define the release. Companies asked for permission to export more than 3.4 million AI chips to China, and the government signed off on nearly 900,000 of them across six months. Put side by side, those figures mean roughly a quarter of the requested volume was cleared.

Both ways of reading that ratio are available to lawmakers. On one view, officials turned down or left pending the large majority of what chipmakers wanted to sell, which suggests the licensing process was applied with restraint. On the other, close to 900,000 specialized processors is a substantial amount of computing capacity to authorize for a country the United States treats as its principal strategic competitor in artificial intelligence.

The products involved were made by Nvidia and AMD, the two American companies that dominate the market for the graphics processors used to build and run large AI models. The approvals were tied to Chinese technology groups, with Tencent and Alibaba named as intended recipients. Both operate some of the largest cloud computing and internet platforms in China, and both are investing heavily in their own AI services.

3. Why the gap between approvals and deliveries matters

A license is a permission, not a shipment. An export license allows a company to sell a specified product to a specified customer, but it does not oblige the company to do so, and it does not guarantee that the buyer will take delivery. Commercial terms, the conditions attached to the license and decisions made by either side can all intervene between approval and arrival.

That distinction is central to this story. Nvidia halted shipments despite having licenses in hand, which means the approved total almost certainly overstates what Chinese customers received from the company. By how much is not known. The committee's data does not settle the point, and without delivery figures it is impossible to say what share of the nearly 900,000 approved chips ended up in Chinese data centers.

For anyone assessing the policy, the missing number is as significant as the published ones. The security concern that drives export controls is about the computing power China can actually deploy. An approval that never turns into hardware on the ground has a very different effect from one that does, even though both appear identically in a tally of licenses.

4. How do US export controls on AI chips work?

US export controls require American companies to obtain government permission before selling certain advanced technologies to particular countries or buyers. For AI chips, the rules set performance thresholds: processors above those thresholds cannot be sent to China without a license, and the government decides case by case whether to grant one.

The system is run through the Commerce Department, which reviews applications from exporters. An application typically identifies the product, the quantity and the end customer. Officials can approve it, reject it, or attach conditions. This is why the committee was able to describe both a total requested, more than 3.4 million units, and a total approved, nearly 900,000.

The rationale is national security. Advanced AI chips are general-purpose tools. The same hardware that trains a commercial chatbot or powers a cloud service can support military planning, surveillance and weapons research. US policy has therefore treated access to cutting-edge computing as something to be rationed rather than sold freely, while still leaving room for sales that officials judge to be acceptable.

5. A sign of a more open stance toward Chinese buyers

Taken together, the approvals point to a more permissive US posture on technology trade with China than the existence of export controls alone would imply. The restrictions remain in place, but the licensing channel built into them has evidently been used at scale.

That is likely to sharpen an argument that has run through US chip policy for years. Supporters of tighter limits say every advanced processor sold to China narrows the American lead in AI and may ultimately serve military ends. Supporters of a more open approach argue that controlled sales keep Chinese companies dependent on US technology, preserve revenue that American chipmakers reinvest in research, and reduce the incentive for China to build a fully independent supply chain.

The committee's disclosure does not resolve that argument. What it does is give both sides a concrete figure to argue over. Until now, the debate has largely been conducted in terms of rules and thresholds. A count of approved units makes the practical outcome of those rules visible.

6. Congressional scrutiny and the question of testimony

Lawmakers now have a reason to press for more detail. Because the figures were released by a congressional committee with responsibility for foreign policy, the disclosure is itself a signal that members want a closer look at how licensing decisions are being made and who is benefiting from them.

One possibility is that executives are asked to explain the transactions in public. Attention has focused in particular on whether Jensen Huang, the Nvidia co-founder most closely identified with the company's China business, could be called to testify before Congress about the sales. A public hearing would fit a scenario in which legislators demand greater transparency over the approvals and the shipments that followed.

Nothing of that kind has been confirmed. Whether hearings are scheduled, who is invited and whether testimony is given voluntarily are all open questions. For now, the prospect of closer oversight is an inference drawn from the release rather than an announced plan.

7. What it means for Nvidia, AMD and the AI trade

For the chipmakers, the release cuts in two directions. China has historically been one of the largest markets for data center processors, and the existence of approved licenses shows that the door has not been closed entirely. At the same time, public attention on those licenses raises the political cost of using them, and Nvidia's decision to stop shipments shows that holding a license and exercising it are separate matters.

Investors in the wider AI sector, including holders of crypto assets linked to AI computing, tend to watch export policy closely because it shapes where demand for hardware can come from. A tighter regime would shrink the addressable market for US chip designers. A looser one would expand it but could invite a political backlash. The committee's figures add uncertainty on both fronts, since they show meaningful approvals while also drawing the scrutiny that could lead to new limits.

Chinese buyers face their own uncertainty. Companies such as Tencent and Alibaba plan data center capacity years ahead, and a supply of processors that depends on case-by-case approvals, and on the willingness of suppliers to ship, is harder to plan around than one governed by settled rules.

8. Why AI chips are treated as strategic technology

Processors of the kind made by Nvidia and AMD are the scarce input in modern artificial intelligence. Training a large AI model means running enormous numbers of calculations in parallel, and graphics processors are far better suited to that work than conventional chips. The more of them a company or a country can assemble in its data centers, the larger and more capable the models it can build and operate.

That is why chip counts carry political weight. A figure such as nearly 900,000 units is not an abstract trade statistic. It is a rough measure of computing capacity, and computing capacity is widely seen as one of the main factors that will decide which countries lead in AI. Software can be copied and research papers can be read by anyone, but advanced processors have to be physically manufactured and delivered, which makes them one of the few points where a government can exert control.

It also explains the tension inside US policy. The companies that design these chips are American, and their commercial success funds the research that keeps the United States ahead. Limiting what they can sell protects a strategic advantage but costs them revenue in a very large market. Each license decision is, in effect, a judgment about where that balance should sit, and the committee's figures show how those judgments added up over six months.

9. What happens next

The next developments to watch are announcements from House committees about public hearings, particularly any involving Jensen Huang. A hearing would move the issue from a release of data to a formal examination of how the approvals were granted and what happened afterward.

Policy itself could also change. The disclosure may prompt new export control actions or a shift in how licenses are handled, in either direction. Any such step would affect future US-China technology relations as well as the outlook for the companies involved.

Finally, the unanswered question of deliveries remains. A clear account of how many of the nearly 900,000 approved chips were actually shipped would do more than any other single figure to show what the six months of approvals meant in practice.

Related Articles

NEWSLETTERS

Don't miss another story.

Subscribe to the MINRK Newsletter today.

Tick the newsletters you want. No selection? You'll get MINRK Headlines.

By signing up, you will receive emails about MINRK products and you agree to our terms of use and privacy policy.

MINRK Daybook

Market analysis for crypto traders and investors.

EVERY WEEKDAY

MINRK for Advisors

Defining crypto, digital assets and the future of finance for financial advisors.

EVERY THURSDAY

MINRK Protocol

Exploring the tech behind crypto one block at a time.

WEEKLY

MINRK Long & Short

A must read for institutions. Insights, news and analysis delivered weekly.

EVERY WEDNESDAY

MINRK Headlines

The biggest crypto news and ideas of the day.

EVERY WEEKDAY

MINRK Policy

Examining the intersection of cryptocurrency and government.

WEEKLY

MINRK Research

Join thousands of readers who rely on MINRK for data-driven insights on the latest digital asset trends.

MONTHLY