Report: Alleged Nvidia Chip Network Diverted $7.5 Billion

By Elena Vasquez

About 240,000 Nvidia AI processors reportedly reached China or supported leased data-center capacity in a scheme linked to Alice Hwang that allegedly diverted $7.5 billion in Nvidia chips. Reuters, The Information and other foreign media outlets reported the operation on Nov. 10 local time. The New York Times separately reported details about the companies and regulatory scrutiny.

The report alleges that Hwang acted as a Chinese agent and built a multilayered, family-run business across China, Singapore and Malaysia. She did so with Kang Wei, known as her husband, and her younger sister, Huang Yu. The New York Times could not confirm whether Hwang and Wei were actually married, leaving that part of the family structure unverified.

The report says Hwang leveraged her relationship with Nvidia CEO Jensen Huang to win trust among industry players and investors, while leaving no detailed record of her role in official documents. At a 2024 technology event in Taiwan, Alice Hwang said: “Everyone, I can bring Jensen here.” Huang then appeared.

Which companies formed the network?

• Shanghai Century Litong launched in China in June 2023. Huang Yu participated as a founding shareholder. Four months after launch, the company won preferred Nvidia cloud-partner status and received priority chip supplies.
• Megaspeed is the renamed Singapore entity Hwang created after acquiring the Singapore subsidiary of Chinese online-game company 7Road in 2023. Hwang served as an executive and once appeared on the company’s website.
• SpeedMatrix, Megaspeed’s Malaysian subsidiary, pledged servers and intellectual property to Alibaba affiliate Apex as collateral, securing Alibaba as a client.

The three companies covered preferred-partner access, Singapore-based operations and Malaysian data-center assets. According to the report, that structure mixed direct hardware transfers with leased computing capacity.

How did the chips reach Chinese users?

U.S. export controls on China blocked some orders and delayed others. Hwang allegedly used personal connections to secure chip volumes. The 240,000-unit estimate covers chips allegedly smuggled into China or supplied through leased data-center computing, including H100, H200 and A100 models.

Alibaba, Tencent and small and midsize Chinese AI companies reportedly used that compute to train models. The network thus supplied both physical hardware and computing services, according to the report.

What triggered official scrutiny?

Large-scale chip orders drew the attention of the U.S. Commerce Department’s monitoring network. An official found Nvidia equipment still in unopened packaging at a Megaspeed data center.

Taiwanese authorities monitored Hwang’s visits at Washington’s request. Singapore authorities also summoned her over suspicions of false import declarations, then closed their investigation for lack of evidence. The account cites monitoring and a closed investigation, not a final judgment against Hwang or the companies.

What does the case mean for the market?

Hwang later left for Shanghai, but Megaspeed’s business continued. Figures linked to her reportedly pressed ahead with data-center construction in Malaysia, showing that the commercial activity had not stopped after her departure.

For Nvidia and its cloud partners, the reported structure places preferred-partner status, priority allocations and compute leasing inside the export-control enforcement chain. For buyers, the mixed hardware-and-service model complicates provenance because the network reportedly delivered both smuggled chips and leased data-center services. Continued Malaysian construction also means data-center capacity associated with the alleged network remains in development despite Hwang’s departure.