On April 4th, according to a recent report by U.S. tech media The Information, Chinese AI startup DeepSeek is on the verge of releasing its next-generation flagship model, V4, which is expected to run entirely on Huawei's self-developed chips. This move is seen as a significant milestone in China's push for semiconductor self-sufficiency. V4 is anticipated to be released within the coming weeks and will operate exclusively on Huawei chips. To achieve this, DeepSeek has collaborated with Huawei and chip designer Cambricon for several months, rewriting parts of the model's underlying code to ensure compatibility with domestic chips. Notably, DeepSeek has not granted early testing access for V4 to Nvidia this time, instead inviting only domestic chip companies to participate in early optimization, breaking industry norms. Chinese tech giants such as Alibaba, ByteDance, and Tencent have already pre-ordered Huawei's latest Ascend 950PR chips, with total orders reaching hundreds of thousands of units. They plan to deploy V4 through cloud services and integrate it into their own AI applications. The surge in demand has driven up the price of these chips by approximately 20%. V4 utilizes a Mixture-of-Experts (MoE) architecture, boasting a total of approximately 1 trillion parameters, with about 37 billion parameters activated per inference. This allows for low latency while supporting multimodal input for text, images, and code. Last year, DeepSeek's release of its low-cost models V3 and R1 caused a global tech stock downturn, leading to market questions about whether U.S. AI companies needed to spend billions on computing power. Consequently, V4 is attracting significant international attention.