DeepSeek Plans 160,000 Huawei AI Chips for Massive New Data Centre

Chinese artificial intelligence company DeepSeek is reportedly planning to deploy at least 160,000 Huawei AI accelerators at a new data centre being developed in Inner Mongolia. If completed as planned, the project would create one of the largest known clusters of Chinese-made AI chips and represent a major investment in domestic computing infrastructure.

The facility is expected to use Huawei’s next-generation Ascend 950DT processors primarily to run DeepSeek’s AI models, a process known as inference. Reports indicate that DeepSeek does not currently plan to rely on the Huawei chips for the training of its models, with Nvidia hardware continuing to play a role in demanding training workloads. This split highlights the different computing requirements involved in developing and operating large AI systems.

The planned deployment is significant because it comes as Chinese technology companies face restrictions on access to some advanced Nvidia processors. Increasing the use of Huawei hardware could therefore help DeepSeek strengthen its computing independence while contributing to China’s broader effort to develop a domestic AI semiconductor ecosystem. The project could also provide Huawei with a major customer for its latest generation of AI accelerators.

However, delivering 160,000 chips could prove challenging. Huawei’s production capacity for the Ascend 950DT is reportedly limited, partly because of component shortages and strong demand from other customers. As a result, completing the full deployment could take more than a year, meaning the timetable for bringing the facility fully online will depend heavily on Huawei’s ability to increase production.

The planned data centre illustrates how geopolitical tensions and restrictions on advanced semiconductor exports are reshaping the global AI industry. Rather than relying on a single hardware supplier, companies such as DeepSeek are increasingly exploring combinations of domestic and foreign computing technologies. If the Inner Mongolia project moves forward at its reported scale, it could become an important test of whether Chinese-made AI chips can support large-scale AI services while reducing dependence on Nvidia.