Nvidia CEO Jensen Huang told Axios in an exclusive interview published July 22 that America has nothing to fear from Chinese open-source AI models, calling Moonshot's Kimi K3 'excellent' and directly challenging Trump administration officials and US AI labs currently lobbying Washington to shut Chinese models out of the American market entirely. It's an unusually direct public split between the CEO of America's most valuable chipmaker and his own government's more hawkish posture toward Chinese AI.
Huang's argument rests on a demand-expansion thesis he's made before, in slightly different form, around DeepSeek: cheaper, competitive open-weight models bring AI capability to more businesses and individual users, which increases total demand for the chips, data centers and computing infrastructure Nvidia sells -- rather than threatening Nvidia's position by proving frontier performance doesn't require frontier compute spend. That framing is a direct rebuttal to the bear case that briefly triggered a Philadelphia Semiconductor Index selloff when Kimi K3's efficiency claims first landed, as investors questioned whether the compute-scarcity thesis propping up Nvidia's valuation still held.
The more pointed part of Huang's comments targeted the tone of the broader AI-safety and China-AI-threat discourse itself: 'I appreciate that many of us grew up and enjoyed science fiction, but it's not helpful,' he said, adding that 'we scare people with science-fiction stories to the point where AI is so unpopular in the United States, or people are so afraid of it, that they don't actually engage it.' He argued Wall Street 'misunderstood the impact of Kimi' the same way it misread DeepSeek's earlier 'DeepSeek moment' -- treating an efficiency breakthrough as an existential threat rather than a market-expanding development.
The timing makes Huang's position genuinely awkward for the administration: the same week he's telling Axios not to fear Chinese AI, the White House is accusing Moonshot of illegally distilling Anthropic's Fable model and separately accessing export-controlled Nvidia GB300 chips through Thailand-based infrastructure -- accusations that, if substantiated, would validate exactly the kind of concern Huang is publicly dismissing as overblown. Treasury Secretary Scott Bessent's talk of sanctions sits uncomfortably next to Nvidia's own CEO calling the underlying Chinese model 'excellent.'
For chip and AI infrastructure investors, Huang's comments are a reminder that Nvidia's business interests -- maximizing total addressable compute demand -- don't always align neatly with US national-security policy, and that gap is becoming harder for Nvidia to paper over as China-AI tensions escalate on multiple fronts simultaneously. For founders building AI products atop either US or Chinese open-weight models, Huang's framing at least offers some cover that using competitive Chinese models isn't inherently the reputational or security risk Washington's rhetoric implies -- though that cover could evaporate quickly if the export-control and distillation accusations against Moonshot are substantiated.
Watch for: whether the administration responds publicly to Huang's comments, given how directly they cut against Treasury's sanctions rhetoric; whether Nvidia faces any political pressure over its CEO's stance while the company simultaneously navigates export-control compliance questions of its own; and whether Huang's demand-expansion thesis holds up if a real sanctions action against Moonshot materializes and disrupts Nvidia's own China-adjacent revenue.