From Huawei Veteran to LPU Rising Star: Yuanchuan Micro's Endgame Bet on Inference Chips

While everyone is fixated on Nvidia's H100 and B200, a Huawei veteran of 22 years is pondering a question: If the endgame for AI isn't larger-scale training but ubiquitous inference, are current chip architectures fundamentally flawed from the start? It's like during the Gold Rush, while others are looking for bigger shovels, he starts tinkering with a new tool specifically designed to sift gold.