If you’re not in the tech world, you might have been only vaguely aware of Nvidia—and CEO Jensen Huang’s leather jackets—until Huang admitted to having an affair with a woman accused of being a Chinese spy. On November 12, a day after revelations of a second affair, this one with an Nvidia executive, Huang resigned. Here’s why the company’s future could affect the entire AI industry.
What Nvidia does
Nvidia is at the center of the AI boom, as the world’s largest supplier of chips used in AI data centers. By some estimates, the company controls 80% of the market. As of November 3, it was the most valuable company in the world, with a market cap of around $5 trillion. By the morning of November 12, as its stock continued to plummet, that dropped to $4 trillion. (Huang’s own net worth had fallen $7.4 billion by that point.)
Huang and cofounders Chris Malachowsky and Curtis Priem launched the company in 1993. They started making graphic processing units, better known as GPUs, for video games. In 2006, they rolled out a platform that enabled other uses for GPUs. In 2012, researchers used the company’s chips to train AlexNet, a breakthrough AI program that taught computers how to recognize images. By the time that ChatGPT launched, Nvidia’s tech was fundamental to AI development.
A large data center might have tens of thousands of Nvidia GPUs that run calculations to train AI models and generate responses when you type a question into ChatGPT or Claude. Training a new AI model could sometimes involve as many as 100,000 GPUs running at the same time for months. After the model is trained, the GPUs are used for “inference,” or using the trained model to, say, rewrite your email or draw a cartoon of your cat. Inference also requires huge numbers of GPUs as millions of people use AI at the same time. (In the U.S., around 60 million people use AI chatbots every day, according to a Pew survey this summer.)
Nvidia systems also include CPUs (“central processing units”) that coordinate tasks and run operating systems in data centers, along with networking equipment that lets GPUs communicate. Its software coordinates the GPUs.
Huang’s downfall
It’s not clear when Jensen Huang started his affair with Alice Huang (no relation), a woman who U.S. intelligence officials reportedly suspect of being a Chinese spy. The Chinese government has denied that she was a spy.
Alice Huang, also known as Huang Le, was a senior executive at Megaspeed International, a Singpore-based company that became an Nvidia customer. Megaspeed reportedly smuggled Nvidia’s chips into China, bypassing U.S. sanctions. The U.S. first restricted the export of Nvidia’s chips to China in 2022, citing the risk that China could use the chips for military development and to compete on AI. Alice Huang reportedly helped move $7.5 billion worth of Nvidia chips to China.
The New York Times first reported on the relationship and the spy allegations on October 29, and Jensen Huang initially said that it was an ordinary business relationship, not an affair. On November 6, the Financial Times reported that it was a romantic relationship. Two days later, Jensen Huang sent an email to Nvidia employees admitting that it had been an affair.
On November 11, reports emerged that Huang also had an affair with one of his employees at Nvidia, reportedly Kimberly Powell, the company’s vice president of health care. On November 12, as pressure mounted from Nvidia’s board, Huang announced his resignation.
Why Jensen Huang can’t easily be replaced
It’s not simple for someone else to step into Huang’s shoes. He was directly involved in decisions that most executives would delegate, from managing in detail how Nvidia’s products were built to how salespeople talked to customers. A typical CEO might have five or 10 direct reports; Huang had as many as 60 direct reports. He reportedly sent hundreds of emails a day.
He had an unusual level of control—and no plan in place for succession. After Huang resigned, Nvidia CFO Collette Kress was named interim CEO while the board searches for a replacement. It’s not clear what will happen next. But it’s possible that the company’s strategic decisions about products, manufacturing, and customers could slow down. Huang oversaw how Nvidia’s full system was integrated, not just its GPUs. The next leader may not have the same vision for the product. Relationships with customers might deteriorate. There might be internal struggles for control.
A circular business model—and potential fallout for AI companies
Nvidia sells its technology to companies like Microsoft, Amazon, and OpenAI. (OpenAI and others also use it indirectly by leasing space in third-party data centers.) But Nvidia is also investing heavily in the AI ecosystem. It committed $30 billion to OpenAI’s financing round and guaranteed up to $105 billion for an Ohio data center that would serve OpenAI. It sells GPUs to AI infrastructure company CoreWeave, but has also invested in the company, including a $2 billion investment in January. Nvidia also agreed to purchase up to $6.3 billion of CoreWeave’s unused computing capacity over the next several years and has struck similar deals to buy cloud computing services from its customers. Meanwhile, it has partnered with Apollo, BlackRock, and other asset managers to mobilize financing for AI infrastructure.
Anthropic, which had plans to file for its IPO next week, is also tightly entangled with Nvidia. Nvidia committed to invest up to $10 billion in the company in 2025; Anthropic committed to use up to 1 gigawatt of Nvidia computing capacity. Nvidia was reportedly considering billions of dollars in credit guarantees for a new data center for Anthropic. It also had reported plans to invest up to $10 billion in Anthropic’s IPO.
All of this meant Nvidia was helping prop up demand for its own technology, making it harder to gauge how quickly AI would have grown on the strength of the underlying economics alone. Huang personally spearheaded these investments and partnerships, and it’s unclear whether his successor will be as willing to finance Nvidia’s customers. That uncertainty could slow some of the AI industry’s planned expansion.
