DeepSeek is reportedly planning to deploy a minimum of 160,000 Huawei AI accelerators in a new data centre in Inner Mongolia. If this move takes place it would create one of the largest known clusters of Chinese AI chips.
Sources told Bloomberg the AI company intends to use Huawei’s next generation Ascend 950DT chips to run its models.
However, delivery of the order could take more than a year. Huawei reportedly does not currently have enough capacity to meet DeepSeek’s planned order as shortages of components including high-end memory are expected to limit production of the 950DT to the low hundreds of thousands this year, with the vendor also balancing demand from other customers.
DeepSeek has asked Beijing to help persuade Huawei to allocate more chips to the company and quicker, according to Bloomberg.
The sources added DeepSeek does not currently intend to use the 950DT chips for training its models, despite Huawei designing the processors for both training and inference. The startup has previously attempted to train models on Huawei chips but continues to rely on Nvidia accelerators for training.
DeepSeek’s rumoured Inner Mongolia project forms part of a wider push to expand its AI infrastructure. Its plans are reportedly at gigawatt scale, with the facility expected to consume enough electricity to power roughly 750,000 homes. The 160,000 Huawei accelerators would form part of the planned capacity.
Source: Mobile World Live
Image Credit: DeepSeek & ShutterStock
Source: Tahawul Tech