DeepSeek plans to use Chinese-made chips, including Huawei processors, for large-scale AI model training as U.S. export controls constrain access to advanced Nvidia hardware. Founder Liang Wenfeng reportedly described the effort as one of the company's biggest bets, while acknowledging that DeepSeek will continue using Nvidia products as Huawei's supply capacity and performance develop. Huawei is expected to deliver a new batch of training chips as early as the fourth quarter or by the first quarter of next year. DeepSeek is training a 2-trillion-parameter model and intends to develop an 8-trillion-parameter model, supported by about CNY 30 billion in computing infrastructure investment. It is also finalizing a second funding round targeting CNY 50 billion at a valuation of up to CNY 500 billion. Huawei's Ascend 950DT, scheduled for a fourth-quarter launch, is designed for training and some inference workloads, but component shortages remain a risk to chip supply.