How a Chinese AI company worked around US rules to access Nvidia’s top chips

But inside a tall, windowless building in Indonesia’s capital, about 2,300 of those chips are ready to do work for a Chinese AI company.

A Wall Street Journal investigation traced how a chain of deals across several countries got the chips inside the data center, which is wedged between a private school and an upscale apartment complex. A company that arranged the transaction is a subsidiary of a Chinese business on an American trade blacklist.

Despite American rules intended to stop China from accessing the tech industry’s most coveted hardware, there is no evidence to suggest the deals violated U.S. law.

Some former and current U.S. national-security officials say the U.S. should review deals such as the Jakarta one. Nvidia and other tech companies argue for fewer export controls, saying it is better to have the rest of the world hooked on American technology and financing American innovation.

The U.S. and China are in an arms race over AI, which can deliver military and economic advantages. America’s edge is that the world’s top AI chip company, Nvidia, is based in California. Since 2022, China has been barred from buying the most advanced U.S. semiconductors over national-security concerns.

Nvidia’s chief executive, Jensen Huang, recently said Nvidia’s market share in China has fallen to zero from 95% owing to U.S. export restrictions. But Chinese companies and organizations still use Nvidia products.

Some bring the chips physically into China using middlemen. Another increasingly popular workaround, which has been employed in Australia and Malaysia, is renting computing power abroad and bringing data out of China and back—sometimes by packing suitcases with hard drives, the Journal has previously reported.

In the Indonesia case, the Journal was able to trace the chips from start to finish, including the specific entities involved. American technology is being made available to a Chinese company through these four steps.

1) Nvidia sells chips to a partly Chinese-owned business partner

Nvidia sells many chips to partners that build servers for AI computing. One partner that makes these servers is Silicon Valley-based Aivres. Its parent is one-third-owned by Inspur, a Chinese tech company. In 2023, the U.S. government put Inspur on a national-security trade blacklist, saying it worked on military supercomputing.

Nvidia can’t do business with Inspur or some subsidiaries added to the blacklist this year. But the rules don’t cover U.S.-based subsidiaries such as Aivres, which has close ties to Nvidia and co-sponsored Nvidia’s most important corporate event this year.

2) The partner finds a buyer in Indonesia for servers with Nvidia chips

In mid-2024, said people familiar with the matter, Aivres began negotiating a deal with the cloud-computing business of Indosat Ooredoo Hutchison, an Indonesian telecom provider. The Indonesian company bought 32 Nvidia GB200 server racks from Aivres for about $100 million, according to people familiar with the contract. Each server rack contains 72 of Nvidia’s leading-edge Blackwell chips.

At roughly 2,300 chips, the processing power is small compared with the tens or hundreds of thousands chips typically required to develop the most advanced AI systems.

Indosat is a joint venture between Qatari telecom company Ooredoo and Hong Kong-based conglomerate CK Hutchison.

3) The Indonesian cloud-computing business lands a Chinese customer

Indosat bought the servers only after securing a client with Aivres’s help, some of the people said. That customer is Shanghai-based AI startup INF Tech.

INF was founded by Qi Yuan, a professor who heads an AI institute at Shanghai’s prestigious Fudan University. Fudan representatives were part of the deal negotiations, according to the people, but ultimately INF signed the contract for the computing power.

4) The Chinese customer aims to use the chips for finance and health AI

As of October, the servers had been delivered and were being installed, one of the people said. When operational, the person said, they will focus on helping INF train AI for financial applications and science research, such as drug discovery.

Lawyers familiar with export-control rules said so long as the Chinese company isn’t using the chips to help China with military intelligence or weapons of mass destruction, the arrangement comports with U.S. export controls under the Trump administration.

Nonetheless, for those who advocate stricter controls on technology, such arrangements raise concerns. They say Chinese companies, even if they have only commercial projects now, may be recruited to help China’s military via Beijing’s strategy of civil-military fusion.

In its final days, the Biden administration created a rule that would have tightened controls over the sale of advanced U.S. chips to countries such as Indonesia that aren’t in a small group of close U.S. allies. The Trump administration later said it wouldn’t enforce the rule.

The rule would have given the U.S. a chance to scrutinize the customer and why it was buying chips, as well as the exporter—especially if it were on the national-security trade blacklist, known as the “entity list," said Thea Kendler, who helped lead export controls under President Biden.

Now, she said, “the government is pushing it on the companies to do their own due diligence."

Aivres didn’t respond to requests for comment. An Nvidia spokesman said the company’s compliance team evaluates and clears its partners before they receive Nvidia products.

“We support the Trump administration’s vision to secure U.S. AI leadership and create American jobs," the Nvidia spokesman said. “The Biden controls cost taxpayers tens of billions, crippled innovation and ceded ground to foreign rivals."

A spokeswoman for the Commerce Department division that oversees export controls declined to comment.

On its website, the Shanghai tech startup INF says it sells AI products focused on the financial and health industries. It was founded in 2021 by Qi, a Chinese-born American citizen with a Ph.D. from the Massachusetts Institute of Technology. Qi was one of the earliest machine-learning scientists at Alibaba, a Chinese tech giant that has helped fund INF.

The startup is seeking to expand outside China, working with data-center operators in Singapore, Malaysia and Thailand to provide AI development services, according to people familiar with its plans.

Asked for comment about its deal with Indosat, INF said it doesn’t do any research with military applications and complies with U.S. export controls.

Indosat is betting on its new AI-chip business as it diversifies from conventional telecom services. Asked whether he had been approached by Chinese customers, Chief Executive Vikram Sinha said Indosat works with multinational companies.

“Any customer which is outside Indonesia goes through the same regulation, whether it is a U.S. company or a China company. If it clears all the regulations, we support it," he said.

Indosat added in a statement that INF has no physical access to its chips. Indosat said its deal with INF would support AI applications tailored for use in Indonesia and Southeast Asia.

Write to Liza Lin at liza.lin@wsj.com and Stu Woo at Stu.Woo@wsj.com