I didn't expect this. So basically if a US company develops a product entirely with it's Chinese subsidiary it's considered a Chinese product?
From the article, it states that "Nvidia said on Thursday the U.S. government has allowed exports needed to complete the development of its flagship artificial intelligence chip". Therefore it is implying that the development work in China still relies on some kind of US input, otherwise Nvidia wouldn't have to send anything over if it was fully localized. On the surface, it sucks that even if the bulk of the development work is being done in China, from a US export controls perspective it is still relying on some US inputs so therefore it got nailed. With the current situation though even if Nvidia was able to complete design development of the chip in China w/o relying on input from the US, the US could still nail Nvidia by invoking secondary rules on its production at TSMC which use US inputs so it would still fall under export control jurisdiction.
For a US company to develop something in any overseas subsidiary, Chinese or otherwise to be considered a non-us product for export control purposes they would have to be zero US inputs and basically be 100% localized or de-Americanized. So in this example Nvidia's Chinese subsidiary would have to do all the work with either all local inputs or if certain content such as code, software libraries, licenses, etc are required they must come from OUS. But yes, technically there is nothing stopping a US company from developing a product entirely under its Chinese subsidiary that is considered a Chinese product. It just requires replicating the same business operations you have elsewhere in China. It's kind of what ARM did with ARM China, by doing a huge IP transfer to its Chinese subsidiary and basically letting them fork it and run with it. They still required some support from other ARM branches as of a few years ago to my knowledge, but that's the kind of localization efforts you would be looking at.