Jensen Huang’s Fall Is Bad for Nvidia but Good for America

Opinion|Jensen Huang Is Gone. Good.

Following Jensen Huang’s resignation from Nvidia after a stunning sex scandal, it’s reasonable to think the U.S. artificial intelligence industry is in trouble. Nvidia’s chips have been vital to A.I.’s ascendance; the company has directly invested billions in A.I. companies, bolstering their chances of success; and Mr. Huang himself was a cheerleader for the go-faster crowd.

The A.I. industry might be wiser to count this as a blessing. Nvidia will take a hit, but American A.I. is in a stronger position to thrive.

Mr. Huang made Nvidia into a bedrock for American A.I. companies, but he was also a baffling champion for their competitors. No person has fought more effectively to give Chinese A.I. companies access to high-powered chips, despite clear national-security risks in areas such as cyberwarfare. When the United States weighed and enacted export controls on this hardware, Mr. Huang repeatedly and successfully pushed back, to the delight of Chinese A.I. developers. Last year, the Trump administration blocked some Nvidia chip sales to China only to reverse course three months later. The administration also allowed Nvidia to sell its powerful H200 chips in China, with a 25 percent cut going to the U.S. government.

To get his way, Mr. Huang built a strong friendship with President Trump. He attended a $1 million-per-person dinner at Mar-a-Lago and assembled a bully of a lobbying operation in Washington. He became so close to Mr. Trump that he took a phone call from the president while onstage at a September industry summit.

Helped by the deals Mr. Huang set up to make Nvidia chips available to China, upstart Chinese A.I. companies made their American counterparts sweat. The first major shock, even before the repeal of H200 export controls, was the unveiling last year of DeepSeek-R1, a Chinese model that matched some American A.I. capabilities at what appears to have been a fraction of the cost. Another shock was Moonshot AI’s Kimi K3, a Chinese A.I. model released this summer that was found to have bested top American A.I. models in certain coding tasks.

Many experts say Chinese A.I. model capabilities lag behind those of American models by just a handful of months. Chinese models tend to be much less expensive to use and more flexible to customize, making them a major nuisance to Anthropic and OpenAI. In one week in September, companies on OpenRouter (a kind of marketplace for choosing different A.I. models to use) had used Chinese A.I. models for more than 50 percent of their activity.