A $480 million Nvidia (NVDA) chip loan tied to ByteDance is on hold. The US rule behind it takes effect Nov. 19.

ByteDance and Nvidia logos above the words: A $480 million Nvidia chip loan tied to ByteDance is on hold

Key points

  • Lenders pause a $480 million ByteDance chip loan
  • Why the Commerce rule reaches a Malaysian data center
  • When the Commerce rule takes effect
  • Where ByteDance's TikTok feed fits in the rule

Lenders have paused work on a $480 million loan to buy Nvidia (NVDA) chips that would be used by ByteDance, the Chinese owner of TikTok, Bloomberg reported on Nov. 17. Bankers at five lenders evaluating the deal discussed pulling out or not joining the loan over concerns that it could violate a new US rule on remote access to AI chips, according to the report.

The borrower, Singapore-based Aolani Cloud, had been in talks with banks to fund AI chips for a data center in Malaysia, the report said. An Aolani spokesperson said the company had "spent significant time with our banking and financing partners to assess and anticipate regulatory risk ahead of this recent new ruling." The spokesperson added: "We are proactively working with regulators and our financing partners to ensure the project delivery goes ahead in full compliance with the new ruling."

Nvidia shares were down about 1.5% at $153.74 as of 11:42 a.m. ET on Nov. 17, ahead of the company's third-quarter results after the close.

The rule covers data centers that serve Chinese companies

The Commerce Department's Bureau of Industry and Security (BIS) released the rule on Nov. 16. It requires a license for advanced AI chips headed to data centers anywhere in the world that give companies headquartered in Macau or in Country Group D:5, which includes China, access to a significant volume of computing power. The rule sets that line at more than 12 million "aggregate TPP," at one data center or across linked data centers. That equals the combined processing power of about 758 of Nvidia's H100 chips, as our explainer on the remote-access rule shows.

The rule also requires a license to export advanced AI chips to Malaysia, Singapore, Indonesia, Thailand, and Brazil, which BIS says present a higher-than-normal risk of diversion. That requirement doesn't apply to chips going to companies headquartered in the US, or whose ultimate parent company is, if the exporter meets new reporting requirements. BIS will review license applications case by case.

Data centers that were already providing that much computing power to companies headquartered in D:5 countries or Macau must get a license within 90 days of the effective date to keep doing so, the rule says.

The rule is scheduled to take effect on Nov. 19

The rule was filed for public inspection at the Federal Register at 4:15 p.m. ET on Nov. 16, under the title "Addressing Remote Access to Advanced Semiconductors." It is scheduled for publication on Nov. 19, and the rule takes effect on its publication date. Comments are due 60 days after publication, and the published rule will give the exact date. If the rule is published on Nov. 19, 90 days after the effective date would be Feb. 17, 2027.

The rule exempts recommender systems, with limits

The rule's summary says the license requirement "will not apply to 'commercial recommender systems,' such as those used by online retailers and social media companies." The regulatory text makes that exception conditional. It applies when the computing power access is "only being provided" for a commercial recommender system and the provider reports three things to BIS: the name of the company it serves, the total TPP provided to that company, and the due diligence it performed to confirm that use.

The rule defines a commercial recommender system as an AI tool exclusively designed and exclusively used to rank, filter, or suggest products, services, or content to users based on their behavior, transaction history, or item attributes. The definition excludes any system that generates text, images, audio, video, software, or other content, even one that also makes recommendations.

The rule's text isn't fully consistent on this point. A later passage in the rule's explanation says data center operators that were providing remote access used for commercial recommender systems "may continue providing these services for 90 days, but must obtain a license following expiration of this period to continue these services." The summary, by contrast, says the license requirement will not apply to those systems.

ByteDance runs TikTok and Douyin, whose feeds rank videos for each user. The company also builds generative AI models, including those behind its Doubao chatbot. Chips used for generative models could require a license if the rule's conditions are met, including the threshold of more than 12 million aggregate TPP. Chips used only to rank content could qualify for the exemption if the reporting requirements are met. Bloomberg's report doesn't establish how the Malaysian project's chips would be used or whether the project would qualify for the exemption.

China has warned it could retaliate

China warned the US on Nov. 17 that it would retaliate if Washington proceeds with curbs on remote access to advanced chips, while urging the Trump administration to resolve differences through dialogue, according to a separate report.

ByteDance has also looked beyond Nvidia for chips. In October, the company was reported to be in advanced talks to buy custom chips from MediaTek for its data centers.

Frequently asked questions

Why did banks pause the $480 million ByteDance chip loan?

Bloomberg reported on November 17, 2026, that bankers at five lenders evaluating the loan discussed pulling out or not joining over concerns it could violate a new US rule on remote access to AI chips. The loan would fund Nvidia chips for a Malaysian data center, through Singapore-based Aolani Cloud, and the hardware would be used by ByteDance.

What does the US remote-access rule require?

It requires a license for advanced AI chips headed to data centers anywhere in the world that give companies headquartered in Macau or Country Group D:5, which includes China, more than 12 million aggregate TPP of computing power, about the combined power of 758 Nvidia H100 chips. It also requires a license to export advanced AI chips to Malaysia, Singapore, Indonesia, Thailand, and Brazil.

When does the US remote-access rule take effect?

On its publication date in the Federal Register, which is scheduled for November 19, 2026. Comments are due 60 days after publication. Data centers already providing that much computing power to companies headquartered in D:5 countries or Macau must get a license within 90 days of the effective date to keep doing so.

Does the remote-access rule apply to TikTok's recommendation feed?

Possibly not. The rule exempts computing power provided only for commercial recommender systems, such as those used by social media companies, if the provider reports the company served, the total TPP provided, and its due diligence to BIS. The exemption excludes any system that generates text, images, video, or other content. One passage in the rule's explanation says data centers serving recommender systems may continue for 90 days but must then get a license, while the rule's summary says the requirement will not apply to those systems.

More on NVDA

Latest stories

Dennis Singleton
Dennis Singleton

Dennis Singleton was born in Australia and later moved to the United States. He has spent years following the markets, but what keeps his attention is how AI is built. He writes about the companies behind the technology, from semiconductor designers and advanced packaging to photonics, memory, networking, and the hardware powering modern AI. His approach starts with filings, earnings, and industry research, then translates the important details into clear, straightforward analysis without unnecessary hype.