A US government agency is conducting a broad review into how Chinese AI companies have continued to access Nvidia's most advanced chips despite years of Washington-led export controls designed to sever that supply line.

According to a Bloomberg report published on August 7, 2026, the review focuses on a workaround that has operated in plain sight: instead of importing restricted chips directly into China — where export controls block such shipments — Chinese firms have been leasing computing power from data centers located in third countries.

The Overseas Rental Loophole

US export restrictions prohibit Nvidia from shipping its most powerful processors directly to China. However, until recently, those same regulations did not prevent Nvidia chips from being deployed in data centers across Southeast Asia, the Middle East, and other regions, where Chinese companies could then lease remote access to that hardware.

Industry estimates indicate that hundreds of thousands of advanced chips may have reached Chinese entities through this regulatory gap before Washington took steps to close it.

The renewed urgency behind the review appears directly linked to recent Chinese AI breakthroughs that demonstrated effective use of Nvidia's restricted processors. Since the first round of comprehensive export controls in October 2022, the US has progressively tightened restrictions — initially prompting Nvidia to develop modified chips such as the A800 and H800 for the Chinese market, which were themselves later brought under control in subsequent updates.

Commerce Department Action Taken in May 2026

On May 31, 2026, the Commerce Department's Bureau of Industry and Security issued updated guidance aimed specifically at closing the overseas subsidiary loophole.

The revised rules changed a pivotal criterion: licensing requirements are now determined by a company's headquarters location rather than the location of its subsidiary. If the parent company is based in China, all entities under its ownership are now subject to the same export controls — regardless of whether they operate in Shanghai, Singapore, or São Paulo.

The guidance explicitly covers Nvidia's most advanced product lines, including the Blackwell and Rubin architectures.

Prior to the May clarification, a Chinese AI firm could establish a subsidiary in a third country, procure chips through that entity, and bypass export restrictions entirely. Under the updated rules, such subsidiaries are treated as extensions of their Chinese parent companies.

Cloud Rental Access Remains Difficult to Regulate

Blocking chip shipments to foreign subsidiaries addresses one part of the problem. Preventing Chinese companies from renting cloud computing time on servers already equipped with Nvidia hardware presents a fundamentally different challenge.

Regulating that type of access would require either restricting who can purchase cloud computing services or mandating that cloud providers implement customer screening protocols tied to end-user nationality. This challenge extends across the major global cloud platforms, where computing capacity is sold through layered reseller and brokerage arrangements that make end-user attribution difficult.

Meanwhile, companies such as Huawei have been developing their own AI processors for years, and tighter US restrictions have only accelerated those efforts. Huawei's Ascend series represents China's most prominent domestic alternative to Nvidia's ecosystem, though industry analysts note that Chinese-developed chips still lag in software compatibility and developer tooling compared to Nvidia's CUDA platform.

Implications for Nvidia and the Broader Market

For Nvidia, the stakes are both regulatory and financial. China was once among the company's largest markets, and each successive round of export controls has further reduced that revenue stream. If the current review results in new restrictions on overseas cloud access, it could diminish global demand for Nvidia hardware by making it less attractive for data center operators that serve Chinese clients.

Companies involved in international chip distribution or cloud computing services may also need to invest significantly in compliance infrastructure, adding costs that are likely to be passed along to customers. The review also signals that US export control enforcement is shifting from a product-level approach — restricting specific chip models — toward an end-user-focused framework that tracks who ultimately benefits from the computing power, regardless of where the hardware physically sits.