240,000 GPUs, $7.5B Claim: What's Proven in the Nvidia China Chip Story

Roughly 240,000 Nvidia H100, H200 and some A100 chips were reported as having moved into China — a figure no regulator has confirmed. A separate $7.5 billion shipment allegation, attributed to The Information and relayed by Reuters on Nov. 10, 2026, is exactly that: an allegation. What does this change for the AI industry? Potentially a lot, because the scandal has exposed a gap in US export controls that lets Chinese firms rent high-end compute without breaking any obvious rule.

The story runs on two tracks, and each feeds the other. CEO Jensen Huang told employees by email he had been "unfaithful to my wife," with reporting connecting the affair to Alice Huang, also named as Huang Le, a businesswoman associated with Singapore-based cloud operator Megaspeed International. The second track is structural. On Nov. 14, 2026, CNN Business reported that the fallout has lifted the curtain on a loophole giving China access to US chips, plus an alleged second channel: physical diversion (buying restricted chips through an intermediary and moving them to a barred buyer). The US probe into Megaspeed has not concluded. Megaspeed and related parties deny any link to illegal diversion.

What is proven, and what isn't?

A short list holds up directly. Megaspeed is a Singapore-based cloud-computing company and an Nvidia customer. Alice Huang is associated with it. A Malaysian subsidiary, Speedmatrix, appears in reporting on the regional network. Beyond that, qualifiers dominate.

• 240,000 chips into China — reported by The Information and others, not independently established, with no breakdown between legal and illegal routes.
• At least $7.5 billion in shipments — unconfirmed allegation; reporting doesn't say whether it covers chips, equipment or services.
• Alice Huang as a Chinese intelligence asset — unconfirmed, denied by Beijing.
• Alibaba or Tencent knowingly joining illegal diversion — unconfirmed; both appear in reporting as end users, which is not evidence of knowledge.

Nvidia said its internal investigation found no information indicating Huang exchanged confidential business secrets with Alice Huang. That statement is narrow. It says nothing about customer compliance with export rules, and nothing about how chips traveled from authorized buyers to Chinese end users.

How does the loophole work?

US export controls target hardware crossing borders. A Chinese company doesn't need to import a GPU if it can rent time on one sitting in a data center in Singapore or Malaysia. The chip never enters China. The compute still serves a Chinese customer. Reporting treats this as a gap, not a clear violation. Physical diversion is different: moving restricted chips into China without a license is a violation, and that's where the legal risk in the Megaspeed probe sits.

Chinese firms effectively have four routes: remote rental of overseas GPUs (legal grey zone, low visibility), physical diversion (illegal if proven), domestic accelerators (legal, weaker performance), and licensed purchase of China-cleared chips (legal, weaker hardware). Remote access combines strong hardware with low regulatory visibility — which is why it draws the heat.

One company gives the gap a concrete shape. Shanghai Century Litong was reportedly founded in June 2023 and held Nvidia "preferred cloud partner" status by October 2023 — roughly four months. Speed proves nothing, but it raises a due-diligence question Nvidia has not publicly answered.

What have the parties said?

Huang's apology, relayed by AFP: "I was unfaithful to my wife. My actions hurt her and our children. There is no excuse for what I did." In a Bloomberg interview quoted by Business Insider, he said: "Our relationship, our interactions, have been related to Megaspeed, and it was all completely normal and not unusual." When Nvidia announced an independent investigation on Nov. 8-9, per CNBC, Huang said: "I support the investigation and will cooperate fully."

Megaspeed, via Reuters on Nov. 12: "Megaspeed takes export control laws seriously and has cooperated with U.S. inspections." Beijing called the spying claim a groundless accusation, slander and a smear.

Who carries the market risk?

Nvidia faces the most direct exposure. Its data-center revenue concentrates in a few big buyers, and any leak in the channel invites stricter reporting rules, tighter end-use checks, and possible limits on who buys the newest parts. Southeast Asian cloud providers could face licensing or screening requirements that smaller operators struggle to afford. Chinese AI labs relying on rented overseas compute carry the biggest strategic risk — a rule change could cut access fast, and smaller labs may have to shrink training runs or delay releases. Even legitimate GPU renters in the region should expect slower onboarding and more compliance paperwork.

The takeaway: the affair supplied the headline, but the loophole supplies the lasting question. Whoever answers it shapes who trains the next generation of AI models.