NVIDIA is currently in the process of developing a tailored B30 AI chip for the Chinese market, marking a significant advancement by supporting multi-GPU expansion for the very first time. This innovation allows users to construct high-performance computing clusters by seamlessly interconnecting multiple sets of chips. The B30 is anticipated to adopt the advanced Blackwell architecture and will be equipped with GDDR7 video memory, deviating from the use of high-bandwidth memory (HBM). Furthermore, it will forego TSMC's cutting-edge packaging technology.