Recently, there have been reports in the market that Huawei’s AI chip is shifting from a dedicated ASIC architecture to a general-purpose GPGPU architecture in order to challenge NVIDIA’s market position. It is said that the new architecture is compatible with the CUDA ecosystem, which can break through existing software bottlenecks and expand more application scenarios. This news has stirred quite a discussion in the tech community. Many articles are comparing the advantages and disadvantages of ASIC and GPGPU, and there is also a lot of commentary on Huawei’s strategic shift…