The Nvidia intelligence scandal now includes a $7.5bn chip-distribution channel
The Information reports that Nvidia CEO Jensen Huang’s relationship with Alice Huang, a Chinese businesswoman whom US authorities reportedly assessed to be an intelligence asset, helped a network of companies obtain at least $7.5 billion of Nvidia chips for Chinese customers. Alice Huang and Jensen Huang are not related. The commercial relationship allegedly dates to at least 2023, when Shanghai Century Litong rapidly gained preferred-cloud-partner status with Nvidia.
That designation mattered because it moved the company closer to the front of a constrained supply chain. When export-controlled H100 and H200 orders stalled, Alice Huang reportedly used relationships at Nvidia and its distributors to obtain supply through companies in China and Singapore. The network acquired approximately 240,000 H100, H200 and A100 accelerators, according to people cited by The Information, with Chinese technology companies and AI laboratories among the ultimate users.
The new allegation changes the risk from a vague counter-intelligence concern into a distribution-control problem. Export controls depend not only on chip specifications and end-user paperwork, but on whether manufacturers can identify who ultimately operates hardware sold through cloud partners and intermediaries. Preferred status, distributor discretion and overseas cloud capacity can each create distance between Nvidia’s sale and the final user. A personal channel would make that system easier to navigate without proving that Jensen Huang knowingly helped evade the rules.
Important facts remain unresolved. Reuters could not independently verify the US intelligence assessment; China rejects it, and Alice Huang has not publicly responded. Nvidia says it has no reason to believe national-security information or trade secrets were compromised. The company’s independent directors have hired outside counsel, while the FBI is also examining whether Jensen Huang was targeted. Huang has acknowledged an affair and promised to cooperate, but that admission does not establish espionage or illegal exports.
The immediate corporate cost is already material: Nvidia lost roughly $430 billion of market value over two trading sessions as investors priced governance, export and succession risk. The decisive evidence will be transactional—partner approvals, chip serials, invoices, cloud deployments and communications—not the sensational personal narrative. If the reported volumes are substantiated, the central failure is that Nvidia’s channel controls could allocate scarce, restricted hardware without reliably constraining its destination.