Nvidia Will Fill Roughly 1% of $70 Billion in Chinese AI Chip Orders

Nvidia plans to deliver about 60,000 RTX Pro 5500 chips against Chinese orders totaling roughly 6 million units — roughly 1% of a $70 billion order book. Bloomberg reported the figure on the 4th, citing supply plans covering requests submitted over four months by Alibaba, Tencent and ByteDance.

Each unit carries a price above $11,000, which is how six million chips add up to $70 billion. Chinese customers had expected to secure about 700,000 units by the end of 2027. That plan has effectively collapsed.

Why can China not simply buy the chips?

U.S. export controls are the constraint. Washington has barred Nvidia from selling its most advanced AI chips to China since 2022. The RTX Pro 5500 sits below the restriction threshold and is cleared for export, which makes it the top-tier AI chip legally available in the Chinese market.

Nvidia has lobbied the U.S. government for years to loosen the rules. It has not won the argument, and with global capacity stretched, the company is steering output toward high-performance, high-margin products for customers such as OpenAI and Anthropic. A company statement said U.S. firms "continue to be constrained by outdated U.S. export controls that cover even gaming products nearly five years old, compounded by China's own restrictions on imports of American-made goods."

Nvidia has also told customers that supply will stay limited for the time being because of product scheduling conflicts. As Nvidia and TSMC allocate more capacity to next-generation products such as Rubin, the RTX Pro 5500 shortage is expected to persist.

What is Nvidia doing about workarounds?

The supply squeeze runs alongside a separate effort to close off indirect routes. The Financial Times reported on the 3rd of last month that Nvidia instructed data center customers to scale back business tied to Chinese clients. Under the new rules:

• Cloud providers buying Nvidia chips must sell at least 51% of their services to companies outside China.
• Chinese customers are capped at 49% of their business.
• The rules cover new contracts only, but apply to Nvidia partners worldwide.
• The guidance was conveyed verbally through sales and regional organizations, not in writing.

Chinese companies have responded to direct purchase limits by renting capacity at offshore data centers in Singapore and Malaysia to train AI models. The new guidance targets exactly those routes. Nvidia neither confirmed nor denied the report.

How far has Nvidia's China position fallen?

The market share numbers are stark. According to IDC, Nvidia held 34% of China's data center AI chip market in the first half of 2026, down from a position where it effectively controlled 95% of the country's advanced AI chip market. China once accounted for about 13% of Nvidia's total revenue.

The order book itself is the clearest measure of China's computing shortage. It also exposes the limits of domestic hardware. Huawei and Alibaba have made real technical progress, but they still trail U.S. competitors in production capability. A prolonged chip shortage would make catching up harder, not easier.

What happens to Chinese buyers now?

Expect further reliance on domestic inference chips from Huawei, Cambricon and T-Head. That substitution carries its own costs. U.S. restrictions on advanced equipment and services, plus a prolonged shortage of high-bandwidth memory, are squeezing Chinese production capacity and pushing prices upward. Cambricon disclosed production constraints in its earnings. Huawei, China's largest AI chip supplier, raised prices on its top-tier accelerators by 60% around September.

The market expects Nvidia's China revenue share to fall further in the near term. Some analysts argue that concentrating capacity on high-margin products is favorable for overall profitability. For China, domestic substitution will accelerate, but capacity and memory bottlenecks look set to hold it back — a shortage on both sides of the Pacific, with no buyer holding the leverage.