TheBizNews

Business news, plainly told.

Read The Brief

US export controls on AI chips face loophole through Southeast Asia rentals

Former Cisco CEO John Chambers recently voiced concerns about the erosion of trust in US technology ties with China, stating he has withdrawn from the Chinese market. His remarks highlight a growing tension surrounding Washington’s efforts to restrict advanced AI chips from reaching Chinese labs.

Hands holding a glowing smartphone in bed at night
Illustration AI-generated (Google Gemini). Not a photo: it does not show the people, places or events in this story. © thebiz-news.com.

Nvidia, a leading AI chip designer, has seen its stock price decline significantly recently, trading at $160.56 on Thursday, down 7.6% for the day and 29.1% over the past month. The company’s leadership is also in flux following the resignation of its CEO.

One significant challenge to US export controls has emerged: a loophole involving data center rentals in Southeast Asia. Companies are constructing data centers in countries like Malaysia and leasing them to Chinese AI firms, including DeepSeek. As reported by Bloomberg’s Odd Lots, this practice is currently legal, with no restrictions on selling chips to Southeast Asia for use in training AI models.

This practice is particularly prevalent in Johor, Malaysia, which has become a crucial hub for Chinese compute power. Nvidia chips are reportedly filling these data centers, many of which wouldn’t exist without the demand from Chinese companies. Huawei, a Chinese competitor, currently produces only a few hundred thousand of its best AI chips annually, with executives estimating it will take until 2030 to meet domestic demand.

Nvidia itself is adapting to this situation, implementing a revenue-sharing model with regional cloud providers like YTL AI Cloud in Malaysia. The company receives payment both on the initial hardware sale and a share of the rental revenue. While this model doesn’t explicitly involve leasing equipment into China, the distributed ownership and third-party operation complicate tracking the ultimate users of the chips.

Nvidia’s recent earnings release indicated that shipments to China-based customers accounted for less than 1% of its data center revenue, which reached $89.02 billion, up 117% year-over-year. The company’s CFO, Colette Kress, highlighted the growth of sovereign AI sales through regional NeoClouds, which more than tripled in the second quarter.

To address this loophole, Washington may consider reviving the AI Diffusion rule or imposing licensing requirements on overseas cloud operators. Monitoring Nvidia’s reliance on buyers in Southeast Asia will also be crucial, as sovereign AI sales through regional NeoClouds represent a significant growth area. John Chambers advocates for a selective approach to advanced technologies, emphasizing the importance of trust between Washington, its allies, and cloud operators.

Source

24/7 Wall St.Original report ·

“Export Controls Were Supposed to Keep AI Chips Out of China. The Rental Loophole Shows Why They Haven’t”

Read the original report ↗

This story was written with the help of AI from the original report credited above and published after automatic checks. No person read it before publication. Errors reported to us are corrected and listed on our corrections page. Read our editorial policy · Report an error · Corrections

Read next

Related

Caught up in five lines

The Brief: the day’s biggest business stories, written by our desk at and .

Read The Brief
Data center aisle lined with glowing server racksUp nextFundstrat’s Tom Lee remains bullish on tech despite Nvidia CEO departure