DeepSeek Plans 160,000 Huawei AI Chips for New Data Centre
Chinese AI company reportedly prepares one of the largest Huawei-powered AI clusters as it seeks to reduce reliance on Nvidia hardware
DeepSeek is reportedly planning to deploy at least 160,000 Huawei Ascend 950DT AI accelerators at a large data centre being developed in Inner Mongolia. If completed, the project would create one of the largest known clusters of Huawei AI chips and represent a significant step in China’s effort to expand domestic computing capacity and reduce dependence on Nvidia.
The planned facility is expected to use the Huawei chips primarily for AI inference, the process of running already-trained models and generating responses for users. Reports indicate that DeepSeek does not currently plan to use the 950DT accelerators for its main model-training workloads, continuing to rely on Nvidia hardware for that more demanding stage of development.
However, fulfilling an order of this size could take considerable time. Huawei’s production capacity for the next-generation chips is reportedly constrained by shortages of components, including advanced memory. Sources cited in the reporting suggest that DeepSeek could face a wait of more than a year for the full quantity, while Huawei balances production among multiple customers.
The project also reflects the growing strategic importance of domestic AI hardware in China. U.S. export restrictions have limited Chinese companies’ access to some advanced Nvidia processors, encouraging technology firms to develop and deploy alternatives from domestic manufacturers. Huawei has invested heavily in AI chips and computing infrastructure as part of the broader push toward greater technological self-reliance.
If the project proceeds as reported, DeepSeek’s planned Huawei cluster could become an important milestone for China’s AI infrastructure industry. It would demonstrate that large-scale domestic accelerators can support the inference demands of a major AI platform, while also highlighting the remaining challenges around chip production, advanced memory and computing capacity. For now, however, the 160,000-chip deployment remains a reported plan rather than a completed installation.
