엔비디아는 미·중 기술 경쟁이 심화되는 가운데 중국의 AI 컴퓨팅 접근을 제한하는 방안을 추진하고 있다.

American semiconductor giant Nvidia is reportedly moving to restrict Chinese technology companies' access to advanced artificial intelligence computing power through overseas data centers, a step that could significantly widen the technology rivalry between Washington and Beijing.

According to a Nov. 3 Financial Times report, Nvidia has instructed its global cloud computing partners to ensure that at least 51 percent of their services are sold to non-Chinese customers. The reported measure would limit the ability of Chinese AI developers to obtain computing power through foreign data centers, an increasingly important alternative to buying advanced American semiconductors directly.

The development comes as the United States seeks to maintain its technological advantage in artificial intelligence, while Chinese companies pursue increasingly sophisticated AI systems despite restrictions on advanced chips.

**Nvidia moves to limit Chinese access to AI computing**

The Financial Times reported that Nvidia began communicating the new requirements to cloud computing providers over the summer, citing five people familiar with the matter.

The instructions reportedly apply to data centers worldwide, not only to facilities in Asia. They affect future contracts but do not alter existing agreements.

Nvidia's sales representatives and regional teams have communicated the restrictions verbally rather than through formal written policies, according to the report.

However, a separate Nov. 3 investigation by The Information, citing three people familiar with the discussions, indicated that the guidance is not currently mandatory. The company has reportedly encouraged cloud providers to diversify their customers and disclose the identities of end users, but has not enforced the 51 percent threshold as a strict requirement.

The distinction leaves open questions about how aggressively Nvidia intends to enforce the reported restrictions and whether they will eventually become binding conditions for its partners.

Responding to the reports, Nvidia neither confirmed nor denied the new measures. In a statement provided to Agence France-Presse, the company emphasized the importance of expanding American AI infrastructure internationally.

"America cannot win the AI race unless America's AI infrastructure wins worldwide," Nvidia said.

The company argued that the global expansion of data centers built around American technology is essential to maintaining U.S. competitiveness in artificial intelligence.

Beijing responded cautiously. During a Nov. 3 press conference, Chinese Foreign Ministry spokesperson Mao Ning said she was unfamiliar with the reported restrictions.

"We believe that all parties should engage in cooperation based on the principles of fairness, transparency and non-discrimination," Mao said, according to the ministry's official transcript.

The reported Nvidia guidance adds another dimension to a technology confrontation that has increasingly moved beyond conventional semiconductor trade.

**Overseas cloud computing becomes a new battleground**

Since 2022, Washington has imposed restrictions on exports of advanced AI semiconductors to China, arguing that the technology could strengthen Beijing's military and intelligence capabilities.

However, limiting physical shipments has not necessarily prevented Chinese developers from accessing American computing technology. Instead of purchasing advanced chips directly, Chinese companies can rent computing capacity from data centers located in other countries. The processors remain overseas, while developers access their computational resources remotely through cloud services.

This arrangement has complicated American attempts to control the international availability of advanced AI hardware. An April 2026 Congressional Research Service report identified remote cloud access as a significant challenge for U.S. export controls. The report noted that restrictions on transferring physical chips do not automatically prohibit every form of remote access, although existing rules can apply to certain end users and sensitive applications.

American lawmakers have attempted to address the issue through legislation. In January, the House of Representatives passed the Remote Access Security Act, which would expand federal authority to regulate foreign access to controlled American technologies through cloud computing.

"The CCP's AI ambitions are being fueled by its access to American chips housed in data centers located outside of China," Rep. John Moolenaar said in a statement announcing the bill's passage.

The legislation reflects growing concern in Washington that traditional export controls may be insufficient in an industry where computing resources can be accessed internationally without transferring ownership of the underlying hardware.

At the same time, U.S. policy has not amounted to a complete prohibition on American chip sales to China. In January, the Department of Commerce revised its licensing policy, allowing certain Nvidia H200 and comparable processors to be considered for export to approved Chinese customers on a case-by-case basis.

"Export controls should evolve with changes in technology, while protecting national security," said Jeffrey Kessler, under secretary of commerce for industry and security.

The competing priorities illustrate Washington's difficult balancing act: preventing sensitive technology from strengthening potential adversaries while preserving American leadership in the global semiconductor and cloud computing markets.

Nvidia's reported restrictions could also create commercial challenges for overseas data center operators, particularly those in Asia that depend heavily on Chinese customers. The Financial Times reported that some providers have struggled to attract enough non-Chinese customers to meet the proposed threshold. Stricter enforcement could force certain operators to reduce their operations or reconsider expansion plans.

**Taiwan crackdown highlights expanding struggle over AI chips**

The reported Nvidia restrictions coincide with renewed efforts in Taiwan to prevent advanced American semiconductor technology from reaching mainland China.

On Nov. 3, Taiwan News reported that Taiwanese lawmakers were considering amendments to export-control legislation that would impose tougher penalties for illegally transferring advanced semiconductors and related technology. The proposed measures could introduce prison sentences of up to 10 years and fines reaching NT$10 million, approximately $304,500.

The legislation follows an investigation in which Taiwanese prosecutors indicted nine people in August for allegedly diverting 74 servers equipped with advanced Nvidia processors to mainland China. The defendants included employees associated with Nvidia and server manufacturer Supermicro.

The proposed restrictions would also target attempts to circumvent export controls by routing semiconductor technology through third countries.

"Changes in supply chains and semiconductor security and resilience are happening faster," Wang Guo-chen, a researcher at Taiwan's Chung-Hua Institution for Economic Research, told Taiwan's Public Television Service.

The case is part of a broader pattern of investigations into suspected attempts to circumvent American restrictions on advanced AI hardware. In March, the U.S. Department of Justice announced charges against three individuals accused of participating in a scheme involving approximately $2.5 billion worth of advanced servers purchased from an American manufacturer. Prosecutors alleged that the defendants used false documentation, intermediary companies and elaborate shipping arrangements to conceal the movement of restricted computing equipment to China.

"These chips are the product of American ingenuity, and NSD will continue to enforce our export-control laws to protect that advantage," said Assistant Attorney General John Eisenberg.

The allegations remain subject to judicial proceedings.

Together, the investigations, proposed legislation and Nvidia's reported cloud restrictions illustrate how the technology confrontation between Washington and Beijing is expanding across