DeepSeek reportedly to utilize Huawei processors for V5 model training
Artificial intelligence developer DeepSeek is planning to shift the training of its upcoming V5 model to Huawei's Ascend chips, according to industry reports.
DeepSeek intends to leverage Huawei Ascend chips to train its next-generation V5 artificial intelligence model, according to a report from the STAR Market Daily published on November 3. The move marks a significant expansion of the partnership, moving from the current use of the hardware for inference tasks to the more demanding process of model training.
This transition signals a deeper integration of domestic hardware within China's artificial intelligence sector. While Huawei's processors have previously supported the deployment of existing models, their application in training infrastructure demonstrates a technical pivot toward domestic computing solutions amidst international export restrictions on advanced semiconductors.
The adoption of Ascend chips represents a strategic alignment for DeepSeek, which has been systematically adjusting its software architecture to ensure compatibility with Chinese-manufactured processors. The industry now awaits further details on how the performance of the V5 model will scale using this hardware configuration compared to previous training cycles.