Jensen Huang used his platform on July 22 to directly counter the panic that's been building around Chinese open-weight AI models, according to Axios's exclusive report. The context: Moonshot AI's Kimi K3 model helped trigger what multiple outlets described as a bear market in the Philadelphia Semiconductor Index, by demonstrating that frontier-adjacent performance doesn't require the frontier-level compute spend that's been justifying Nvidia and AMD's valuations.
Huang's position -- that competitive Chinese AI, including open-weight models, ultimately grows the total pool of AI compute demand rather than shrinking Nvidia's slice of it -- is consistent with arguments he's made before, but the timing matters. It comes the same week Treasury Secretary Bessent said the US could pursue sanctions against China over alleged AI model 'theft,' and amid a broader Washington mood that's grown more hawkish on Chinese AI capability generally, including proposed restrictions tied to Trump's push for American-made AI chips that CNBC reported is already squeezing TSMC's margins.
“Huang is effectively staking out a contrarian position relative to his own government's posture, and relative to the market's initial reaction to Kimi K3.”
Huang is effectively staking out a contrarian position relative to his own government's posture, and relative to the market's initial reaction to Kimi K3. Nvidia's fortunes are more tied to global AI compute demand than to any single lab's dominance, which gives Huang a real incentive to talk the market off the ledge -- more AI adoption anywhere, including inference on cheaper Chinese open-weight models, still needs chips to run on, even if it's not always Nvidia's most expensive parts.
The broader industry is visibly recalibrating. The Register's companion piece -- 'the truth nobody wants to admit: Chinese or not, open models are competitive now' -- captures a real shift in sentiment that goes beyond one company's earnings reaction. DeepSeek, Moonshot and other Chinese labs have spent the year proving that efficiency gains can substitute for raw compute scale, which is exactly the finding that spooked chip investors when Kimi K3 landed.
For VCs underwriting AI infrastructure and chip-adjacent bets, Huang's comments are a signal worth weighing against the more bearish read from the semiconductor selloff itself -- valuations built on scarcity of compute are more fragile than valuations built on total AI adoption growth, and founders pitching 'more compute demand' stories should be ready to defend that thesis against efficiency-driven Chinese competition specifically, not just hand-wave at 'AI is growing.'