Jensen Huang, Mega Speed, and 600,000 Stranded Nvidia Chips Explained
Nvidia's board is probing Jensen Huang's ties to a suspected Chinese intelligence contact as chip export questions and the Mega Speed case collide.

What is the Jensen Huang scandal actually about?
Nvidia’s board is investigating whether CEO Jensen Huang’s personal relationship with an executive, who US officials reportedly suspected of ties to Chinese intelligence, had any bearing on who got access to Nvidia’s advanced chips. The scandal has two threads that keep crossing: Huang’s acknowledged affair, disclosed to employees on November 9, and a separate, longer-running federal investigation into whether a Singapore-based firm called Mega Speed diverted restricted Nvidia hardware toward China. Neither thread proves wrongdoing on its own. Together they’ve triggered a board inquiry, congressional attention, and new scrutiny of how US chip export policy actually gets made.
TL;DR
- Jensen Huang told Nvidia staff on November 9 that he’d been unfaithful to his wife, without naming the other person, and Nvidia’s board immediately opened an independent investigation into the relationship.
- US officials reportedly believed the woman involved had a connection to Chinese intelligence, and the New York Times reports Donald Trump was briefed on that suspected link in October of last year, a full year before the affair became public.
- About 600,000 Nvidia H200 chips are sitting in Hong Kong, according to Reuters, legally approved for sale by Washington but held up by Beijing, which reportedly wants to reduce dependence on American chip suppliers.
- Mega Speed, a Singapore company that buys Nvidia hardware and rents computing power across Southeast Asia to clients including Alibaba, has been under US investigation since 2024 over whether chips were illegally routed into mainland China.
- Bloomberg’s analysis of 2025 trade records found Mega Speed importing more than 17,000 GB300 processors in a single month, among the most advanced systems Nvidia makes, even though Nvidia says the overseas rental business is permitted under export rules.
- Nvidia’s top lawyer contacted Commerce Department investigators in August 2025 citing authorization from Commerce Secretary Howard Lutnick, a contact some officials have called unusual, and the case has since moved to a Department of Justice criminal inquiry.
- Congress has asked Huang to accept a closed-door briefing around November 18, one day after Nvidia’s quarterly earnings call, putting both the affair and the export-control questions in front of lawmakers within the same week.

Nothing gets lost.
How did the affair and the chip investigation become one story?
They started as separate issues and converged through timing and personnel. The Mega Speed investigation into possible chip diversion has been running since 2024, well before the affair became public. Separately, US officials reportedly flagged a suspected Chinese intelligence connection tied to an executive in Huang’s orbit, information serious enough that Trump was briefed on it in October of last year, according to the Times. Two months after that briefing, Trump opened the door for Nvidia to sell China the H200, a more powerful chip than had previously been permitted for export there.
That sequence is what turned a personal matter into a policy question. Huang disclosed the affair to Nvidia employees on November 9 without naming the woman. Nvidia’s board responded by launching an independent investigation specifically into whether the relationship influenced decisions about chip access. The board has said it retains confidence in Huang and has found no evidence that Nvidia secrets or national security were compromised. The Times has also reported that officials initially believed Huang did not know about the woman’s alleged intelligence ties and had not received a defensive briefing warning him about the relationship. China has denied the espionage allegations.
What is Mega Speed and why does it matter?
Mega Speed is a Singapore-based company that buys Nvidia equipment and rents out computing capacity from data centers across Southeast Asia to outside customers, including Alibaba. The arrangement lets a Chinese company use Nvidia’s most advanced hardware for training or running AI models without the physical chips ever entering China. Nvidia says this overseas cloud rental model is permitted under current export rules.
The investigation centers on a narrower question: whether some of the physical chips tied to Mega Speed’s operations were also moved into China through other channels, which would violate export controls. Bloomberg’s review of 2025 trade records found Mega Speed importing more than 17,000 GB300 processors, built on Nvidia’s newer Blackwell Ultra architecture, in a single month. That’s a meaningful volume of frontier-class hardware flowing through one company’s supply chain.
Nvidia disputes the diversion allegations and says its internal compliance checks have found no evidence of smuggling. Mega Speed says it has repeatedly offered to cooperate with investigators. The Commerce Department says it followed standard procedure throughout. But the case has escalated: in August 2025, Nvidia’s top lawyer contacted Commerce investigators and invoked authorization from Commerce Secretary Howard Lutnick. Investigators reportedly declined a meeting and raised concerns about the contact. The matter has since moved to the Department of Justice, where a criminal inquiry is reportedly ongoing.
Why are 600,000 Nvidia chips stuck in Hong Kong?
Reuters reported on November 5 that roughly 600,000 Nvidia H200 chips were sitting in Hong Kong, cleared by Washington for sale but not moving into mainland China in bulk. Some smaller shipments had gotten through. The holdup, according to Reuters, traces back to Beijing, not Washington. Chinese authorities are reportedly wary of letting domestic AI companies become too dependent on American chip supply, even when US regulators have approved the sale.
This creates an odd dynamic: Chinese firms want more computing power, but their own government is the one slowing the flow of legally available chips. At the same time, the same companies can rent access to even more advanced hardware (like the GB300 systems tied to Mega Speed) sitting in data centers outside China’s borders, which sidesteps the question of direct ownership entirely. A chip installed in Malaysia can still do economically useful work for a Chinese company if that company is renting the compute, even though the hardware itself never crosses into Chinese territory.
Why does chip generation matter this much?
Not all Nvidia chips offer the same capability. The H200 is an older processor still capable of training and running large models. The GB300, built on the newer Blackwell Ultra architecture, offers substantially more capacity for demanding AI workloads at the scale of a serious research lab. The difference affects how fast a company can run experiments, how cheaply it can serve a model to customers, and how close it can get to the frontier of what AI can do.
That’s why access, not just chip count, has become the real currency. A company with early access to the most capable hardware can iterate faster, train bigger models sooner, and extend whatever lead it already has. That’s also why a supplier relationship, and any question about who influences it, carries weight far beyond one company’s balance sheet.
What does this have to do with AI improving itself?
Separately from the chip story, there’s growing evidence that frontier AI labs are using their own models to accelerate the development of the next generation of models. METR, a research organization that evaluates AI capabilities, spent 11 weeks examining Anthropic’s internal data and interviewing staff. Most of the team concluded that Anthropic’s pace of improvement looked roughly twice as fast as it was between spring 2024 and fall 2025, essentially two years of progress compressed into one, with a plausible range of 1.3x to 3x given the uncertainty involved. METR attributed most of that acceleration to Anthropic’s own use of AI in its research process.
OpenAI has reported that experiments per active researcher hit a new high in August, in records going back over a year and a half, driven by heavier use of coding agents and more available compute. At Google, CEO Sundar Pichai has said agents are already helping researchers build models, while cautioning that fully autonomous self-improvement remains some distance off. The pattern across labs looks like assisted recursive self-improvement: humans still choosing which problems to pursue, AI handling more of the coding, testing, and experiment design in between. Access to the most capable chips feeds directly into this loop, because faster, bigger compute means faster iteration on the models doing the accelerating.
What happens next?
Big news
firstNov 17
Congress has asked Huang to accept a closed-door briefing by November 11 or propose an alternate date, with November 18 floated as the target, one day after Nvidia’s quarterly earnings call on November 17. That gives lawmakers and investors back-to-back opportunities to press Huang on decisions stretching back more than a year, including the July 2025 meetings in Washington and Beijing where Nvidia secured assurances on H20 chip licenses.
If the board’s investigation and congressional review turn up no evidence that the relationship influenced chip-access decisions, this likely fades, given how much of the US economy is currently betting on chip supply continuing to flow. If investigators find otherwise, the fallout extends well past Huang’s career, into how quickly both the US and Chinese economies can build on frontier AI.
Frequently Asked Questions
What did Jensen Huang actually admit to?
Huang told Nvidia employees on November 9 that he had been unfaithful to his wife. He did not name the other person involved. Nvidia’s board then opened an independent investigation into whether that relationship affected any business or chip-access decisions.
Who is under suspicion of ties to Chinese intelligence?
The person is unnamed in public reporting. The New York Times reports US officials suspected an intelligence connection tied to an executive associated with Mega Speed, and that Trump was briefed on this suspicion in October of last year.
Is Mega Speed accused of breaking export laws?
Investigators have examined whether chips tied to Mega Speed’s operations were illegally diverted into mainland China. Nvidia disputes this and says its compliance checks found no evidence of smuggling. Mega Speed says it has offered to cooperate. The case has reportedly moved to a Department of Justice criminal inquiry.
Why won’t China take chips it’s allowed to buy?
Reuters reports Beijing is wary of Chinese companies becoming too dependent on American chip supply, even when Washington has approved the sales. That caution is reportedly behind roughly 600,000 H200 chips sitting unclaimed in Hong Kong.
Does this affect AI progress more broadly?
Yes, indirectly. Whoever controls access to the most advanced chips can train and run more capable models faster. Combined with early signs that AI is already helping accelerate its own research cycle at labs like Anthropic and OpenAI, questions about who gets priority chip access carry weight beyond any single company.
