The US is winning the AI race. Jensen Huang could have blown it

Most coverage of the Nvidia scandal has focused on what Jensen Huang knew, what Nvidia's board will find, and what an intelligence service might have learned. This piece asks a different question: what does it cost the United States in the race itself? Experts told The Verge that China trails the US by a matter of months, which means even modest delays at Nvidia could matter.

Two concrete risks stand out. One is allocation: if companies linked to the alleged spy had priority access to Nvidia's newest GPUs during a shortage, American buyers may have waited longer. The other is distraction: investigations that consume Nvidia's leadership could slow decisions, product rollouts and the allocation of chips that labs such as OpenAI and Anthropic depend on.

We think the most useful frame here is the one CNAS senior fellow Emily Kilcrease offers: China has energy and a strong open-weight model ecosystem, but without leading Nvidia chips it can't close the gap with US frontier capabilities. With them, she says, the lead "could vanish in a matter of weeks or months." If that's right, the chips are the race, and anything that bends who gets them is a strategic event, not a corporate one.

That's why the priority-access question from Brookings fellow Kyle Chan deserves more attention than it's gotten. The Verge reports that companies linked to the alleged spy, Alice Huang, no relation to Jensen, reportedly enabled at least $7 billion of Nvidia chip sales to China. There's no public evidence that Jensen Huang directed restricted chips there or handed over sensitive information, and we want to keep that front and center. But in a shortage, allocation is close to zero-sum. If China-linked buyers really were near the front of the line, US labs were pushed further back. That's a harm worth investigating whether or not any law was broken.

The second risk is quieter and maybe larger. Kilcrease warns the investigations will consume an immense amount of bandwidth inside Nvidia. Switching to other hardware takes time, she notes, so if Nvidia stumbles on execution, labs can't simply pivot. We've argued before that the AI economy leans too hard on one company. This piece makes the cost of that concentration concrete: a slowdown at Nvidia becomes a slowdown for the whole American frontier, measured in exactly the weeks the race supposedly turns on.

We'd push on the race framing a little, though. Treating AI as a sprint where months decide everything can justify cutting corners on safety, export rules and governance, and speed-at-all-costs is the same logic used to argue for loosening chip sales abroad. A durable lead built on secure supply chains and trustworthy institutions is worth more than a few weeks shaved off a training run.

Which brings us to the irony. Huang has been one of the loudest critics of chip export restrictions, and last year won White House approval to export H200s to approved customers in China. Now, Kilcrease says, the odds that Congress passes AI chip control legislation have soared. If Beijing's operation was meant to loosen the flow of chips, exposure may tighten it for years. I'm watching whether that legislation moves, and whether it's written carefully enough to close the gaps without strangling legitimate trade.

The US restricts sales of its most advanced AI chips to China, and a black market for restricted Nvidia chips has emerged. Jamestown Foundation fellow Sunny Cheung told The Verge that the intelligence value of access to Huang extends beyond trade secrets to insight into White House thinking on export controls and executives' private views on China and Taiwan. Nvidia shares have slumped since news of the affair broke.