The Nvidia Scandal Could Upend Chip Sales to China. Trump Wants to Keep Them Going
Washington’s disagreement is over how tightly to restrict advanced Nvidia chips going to China. Treasury secretary Scott Bessent challenged commerce secretary Howard Lutnick for not moving faster to add Megaspeed to the US entity list. Lutnick had waited for Trump’s direction, while other officials supported listing the company. Trump chose a different response: a new interim rule aimed at stopping Chinese AI labs from remotely leasing compute power from data centers in Southeast Asia. The article says Megaspeed already had the chips, so blacklisting it alone could be like “playing whack-a-mole” without addressing remote access. Trump’s stated reasoning is that allowing some sales could keep China reliant on US technology and boost US tech stocks. A senior White House official said China must not get Nvidia’s most advanced Blackwell and Rubin chips. Megaspeed had acquired H200 chips, which use an earlier-generation architecture, so the debate also concerns which chip generations may be sold.
What is the central disagreement in Washington over selling Nvidia’s advanced chips to China, and why does Trump want some sales to continue?
Washington’s disagreement is over how tightly to restrict advanced Nvidia chips going to China. Treasury secretary Scott Bessent challenged commerce secretary Howard Lutnick for not moving faster to add Megaspeed to the US entity list. Lutnick had waited for Trump’s direction, while other officials supported listing the company.
Trump chose a different response: a new interim rule aimed at stopping Chinese AI labs from remotely leasing compute power from data centers in Southeast Asia. The article says Megaspeed already had the chips, so blacklisting it alone could be like “playing whack-a-mole” without addressing remote access.
Trump’s stated reasoning is that allowing some sales could keep China reliant on US technology and boost US tech stocks. A senior White House official said China must not get Nvidia’s most advanced Blackwell and Rubin chips. Megaspeed had acquired H200 chips, which use an earlier-generation architecture, so the debate also concerns which chip generations may be sold.
What is the US entity list, and what restrictions can apply to companies placed on it?
The article describes the entity list as a tool for restricting trade with named companies, but it does not explain the list’s rules. As general background, it is a US Commerce Department list of foreign persons and organizations considered a national-security or foreign-policy concern. Its purpose is to control access to certain US goods, software, and technology.
In general, a company on the list may face licensing requirements for exports, reexports, or transfers of items covered by US export regulations. The specific limits depend on the listing and the items involved; a license may be denied. This is different from an automatic ban on every possible transaction, so the details matter.
In the article, officials discussed adding Megaspeed to the list because it had received advanced Nvidia chips through a chain of companies. The article also describes a separate “affiliates rule,” which extended restrictions to companies at least 50 percent owned by listed entities. That rule was later suspended after Chinese retaliation and criticism inside the US government.
How did Megaspeed give Chinese companies access to Nvidia chips through data centers in Southeast Asia?
Megaspeed’s arrangement separated the location of the chips from the location of the customers using them. Instead of shipping the physical Nvidia chips into China, Megaspeed hosted servers containing them at data centers in Malaysia and Indonesia. Chinese companies could then rent computing capacity from those servers.
The route to Megaspeed ran through Aivres, a California-based company operating as a subsidiary of Chinese server conglomerate Inspur. The article reports that Aivres could legally buy restricted Nvidia chips without triggering direct export bans, then transfer them internationally to Megaspeed subsidiaries. Aivres said it complies with US export controls.
This setup meant Chinese firms could use the chips’ computing capacity without the hardware being physically located in China. Aivres transferred billions of dollars’ worth of restricted chips, including H200s. The proposed interim rule would target this remote-access route by stopping Chinese AI labs from leasing compute power through data centers in Southeast Asia.
How much were the restricted Nvidia chips that Aivres transferred to Megaspeed’s subsidiaries worth?
The article’s stated scale is “billions of dollars’ worth” of restricted Nvidia chips transferred by Aivres to Megaspeed’s subsidiaries. It does not provide an exact dollar amount, a chip count, or a breakdown by model. The transfer included powerful H200 chips, which use Nvidia’s previous-generation Hopper architecture.
Aivres was a California-based company and a subsidiary of Inspur, the Chinese server conglomerate linked to Megaspeed. The article says Aivres could legally purchase restricted Nvidia chips without triggering direct export bans, then ship them internationally to Southeast Asia. Megaspeed hosted servers there rather than sending the physical chips into China.
Major Chinese companies could rent computing capacity from those servers remotely. That arrangement is central to the policy dispute: the chips were outside China, but their computing power could be accessed by Chinese firms. Trump’s proposed interim rule would target remote leasing, while the article does not state a final value for all transferred chips.
Why would a new rule targeting remote access to chips address a gap that shipment bans might miss?
A shipment restriction can miss a transaction when chips do not physically enter China. In the arrangement described, Megaspeed hosted servers in Malaysia and Indonesia, then rented their computing capacity to major Chinese companies. The hardware stayed outside China, while Chinese users could access the computing power remotely.
That is why Trump settled on a proposed interim trade rule targeting remote access. The rule was intended to stop Chinese AI labs from leasing compute power from data centers in Southeast Asia. It addresses the service created by the chips, not only the movement of the chips themselves across borders.
The article says Megaspeed had already possessed the chips for months, so adding it to the entity list alone could lead to a “whack-a-mole” response if the underlying access route remained open. The White House said the rule would also crack down on smuggling and direct advanced chips first to US innovators. The article does not say whether the rule has taken effect.
How can tightening chip export controls lead to retaliation in other industries, as happened with rare earth metals?
The article describes a direct example of retaliation crossing industries. In September 2025, the Commerce Department began aggressively applying an affiliates rule that extended restrictions to some companies owned by entities already on US control lists. China responded by suspending exports of critical rare earth metals to the United States.
That response mattered beyond the chip trade because the restrictions concerned companies, while China’s countermeasure affected a different category of goods. The article says the suspension prompted internal criticism in the US, with some accusing the Commerce Department of having sparked a trade war. It does not specify the metals’ uses or quantify the impact.
After the Busan summit at the end of October 2025, Trump and Chinese leader Xi Jinping agreed to a trade truce. The White House then told Commerce to halt use of the affiliates rule and clear new export controls in advance. The sequence shows why officials may weigh the risk of retaliation when considering new restrictions, though the article does not predict what China might do next.
What is computing power, and how can a company rent it remotely without owning the chips that produce it?
Computing power means the capacity of computer hardware to process information and run tasks. The article uses the term for what Chinese companies rented from Megaspeed: access to servers powered by Nvidia chips. This definition is general background; the article describes the arrangement but does not define computing power.
A company does not have to own the hardware to use its capacity. In general, a service provider can keep servers in a data center and let customers connect remotely to run computing workloads. The customer pays for access to the service rather than buying and operating the physical chips itself.
That distinction is central to the article’s account. Megaspeed hosted servers in Malaysia and Indonesia, while major Chinese companies rented their computing power remotely. The chips were not shipped into China, but the companies could still access their capacity. Trump’s proposed rule would target Chinese AI labs leasing that compute power from Southeast Asian data centers.
This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.
Read more in the JupiteX app
Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.
Or read more news on the web