It's not about stopping, it's about delaying.

There's a reason Huang's quote is in context of PRC talent.

PRC generates plurality of global AI talent and is projected to generate disproportionate more short/medium term. Nvidia want to delay future where 50%+ of global AI talent working on Huawei/Ascend ecosystem or fragment PRC talent pool so net 50%+ of global AI talent works on CUDA moat for as long as possible (which is strategicially significant depending if you're near/medium term AGI believer). The reality is huge % of foreseeable US AI talent is going to be from PRC anyway, but without fragmenting domestic PRC talent, PRC will be net talent winner. The second order effects of that is PRC will have single direction valve to siphon knowledge (knowledge diffusion + espionage) from SV but the vice versa will be more controlled, a PRC AI blackbox, and faster PRC proliferates domestic AI the less ability for west/NVIDIA to brain drain PRC talent developed from different system. And with hardware gap closing in 5/10/15 years, and energy gap firmly in PRC favour (i.e. US likely not able to close energy gap faster than PRC can close compute gap), and you basically have trendline where PRC will sprint from a few years behind to years ahead once compute constraint lifts from their talent advantage.