Nvidia's Huang urges US and UK talks with Chinese AI labs on safety

Huang's request costs Nvidia little to make. CNBC describes him as a prominent voice against the regulation that OpenAI's Sam Altman and Anthropic's Dario Amodei have called for ^[6], and Qz reports that he wants the industry to accelerate and says AI can be used to monitor AI ^[11]. Contact with Chinese researchers and monitoring both fit that position (a pause would not). Huang said, "We ought to enhance the dialogue" ^[2], and in our view that extends his existing stance.

The money is the only number in the package. Qz reports that Nvidia is donating $250 million over five years toward UK AI safety research ^[8]. That averages $50 million a year ^[7], delivered as access to time on advanced computers ^[8]. What that time is worth in cash depends on whose computers are used and at what hourly rate. The Qz report does not state either.

On chip policy, we think the near-term signal is small, or rather small that anyone has written down. Huang, Bessent and Xi's readout each name talks as the next step: Huang wants dialogue ^[2], Bessent discussed the prospect of a "U.S.-China AI dialogue" with He Lifeng in September ^[4], and the Chinese readout has Xi saying the two sides "can continue their dialogue on AI" ^[5]. The channel remains something that was discussed.

The counter-thesis is that an incident channel is where hardware questions would surface first, since a misuse report on a particular model invites the question of what it ran on. We would change our view if the channel gets a written scope that names compute or model providers, or if Huang's hypothetical about cutting off a customer ^[10] becomes a stated condition.