DeepSeek is planning to deploy at least 160,000 Huawei Ascend 950DT chips at its AI data center under construction in Ulanqab, Inner Mongolia, according to Bloomberg citing sources. This would create one of the largest known clusters of Huawei AI chips. However, Huawei's production capacity constraints could delay full delivery by over a year. The chips will run DeepSeek models but are not intended for training, which will continue to rely on Nvidia GPUs. The 160,000 chips represent only part of the broader project, which includes a 1GW AI data center and additional rented computing power.
DeepSeek plans to deploy at least 160,000 Huawei Ascend 950DT chips at its AI data center under construction in Ulanqab, Inner Mongolia, according to Bloomberg citing sources familiar with the matter. If realized, this would become one of the largest known clusters of Huawei AI chips to date.
DeepSeek intends to purchase more of these chips, but Huawei's production capacity is currently unable to keep up, with full delivery potentially taking over a year. The 950DT batch will be used to run DeepSeek's models, and the company has no plans to use them for training at this stage. DeepSeek continues to rely primarily on Nvidia GPUs for its core training tasks.
The 160,000 chips are only part of the broader project. DeepSeek had previously planned to build an approximately 1GW AI data center in Ulanqab, and is also renting additional computing power from other local operators to support its expanding AI computing needs.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
