fun with mafs but the actual reality is that a dollar goes *further* in the US. Models are more powerful, chips do more per watt and second and $, more data has already been collected, labs are larger, older. All other inefficiencies return us to roughly the raw value gap.
Have received a few questions on this. How can the US really be spending 7x China on AI?
Part of the answer is that building stuff in China is cheaper. Also the state is providing some of the funding. The "true" gap is more like 2-3x. @hsu_steve @stevehou







