DeepSeek shifts its inference workload to Huawei
DeepSeek plans to install at least 160,000 of Huawei's next-generation Ascend 950DT chips at a new gigawatt-scale data centre in Ulanqab, Inner Mongolia, according to a Bloomberg report cited by The Decoder. The company intends to run only inference workloads on the Huawei chips, keeping Nvidia hardware for training even though the 950DT is designed to handle both. If it is built at that scale, the site would form the largest known cluster of Chinese-made AI accelerators, and it is expected to consume enough electricity to power roughly 750,000 homes. Wccftech, drawing on the same Bloomberg reporting, put the order's value near 2.56 billion dollars, based on the Ascend 950DT's street price of about 16,000 dollars per chip, below the 15,000-to-25,000 dollar range Nvidia's China-specific H20 chip commands. Each Ascend 950DT carries 144 GB of high-bandwidth memory and can be linked with up to 8,192 other units into Huawei's Atlas 950 SuperPoD cluster.[1], [2]
