DeepSeek reportedly plans to deploy at least 160,000 Huawei AI accelerators at a new data centre in Inner Mongolia, potentially creating one of the largest known clusters of Chinese-made AI chips.
According to sources cited by Bloomberg, the AI company intends to use Huawei’s next-generation Ascend 950DT chips to run its models. However, delivery of the chips could take more than a year because Huawei currently lacks sufficient production capacity to fulfil the planned order.
Production of the Ascend 950DT is expected to remain in the low hundreds of thousands this year, partly because of shortages of components, including high-end memory. Huawei is also balancing demand from other customers.
DeepSeek has reportedly asked Beijing to help persuade Huawei to allocate more chips to the company and accelerate deliveries.
Despite the Ascend 950DT being designed for both AI training and inference, DeepSeek does not currently plan to use the chips to train its models. The startup has previously attempted to train models on Huawei chips but continues to rely on Nvidia accelerators for training.
The proposed Inner Mongolia project is part of DeepSeek’s wider effort to expand its AI infrastructure. Its plans are reportedly at gigawatt scale, with the facility expected to consume enough electricity to power about 750,000 homes. The 160,000 Huawei accelerators would form part of that planned capacity.
Huawei outlined a three-year roadmap for its Ascend chip portfolio last year, including the 950 series, with the 950DT scheduled to ship in the fourth quarter of 2026.
At the time, Huawei rotating chair Eric Xu acknowledged that its individual AI chips were less powerful than Nvidia’s, but argued that connecting them into large-scale systems could overcome the performance gap.
Huawei’s expansion of its AI chip business remains constrained by Western sanctions restricting its access to advanced semiconductor technology, while US export controls have limited Nvidia’s access to the Chinese market for advanced AI chips.
