Chip War has a great section on how the Soviet Union tried a “just copy/steal” strategy in semiconductors and fell hopelessly behind because of it. It’s a great theoretical idea to just copy/steal and fast-follow, but semiconductors, AI, and other “harder technologies” require building human and intellectual capital that will get better with time. From there, you need to have the prior generation to keep up with ever-increasing complexity and difficulty as these things get more advanced.

I disagree with your section on Huawei and China. China isn't just trying to just copy/steal AI. In terms of models, China is a bit behind in LLMs but arguably more ahead in self-driving cars. China is throwing everything at semiconductor manufacturing instead because that's where their bottleneck truly is - not CUDA. Had Huawei had access to TSMC's 5nm and 3nm, they might already be equal to Nvidia in raw GPU prowess. After all, HiSilicon's Kirin already matched/exceeded Qualcomm before the Trump ban. Their 5G chips/implementation were well ahead of anyone else. In software, it's easier for China to adopt a CUDA alternative because China is usually really good at unifying under one vision - especially when they have to.