One-sentence summary
CLS reported on November 3, citing multiple supply-chain sources, that DeepSeek's next-generation model V5 is planned to be trained on Huawei Ascend chips, which previously served only DeepSeek's inference workloads.
Event Overview
On November 3, 2026, CLS reported that DeepSeek plans to train its next-generation large model, DeepSeek V5, on Ascend chips, according to multiple industry sources.
Key Information
- Ascend chips have been used mainly for DeepSeek's inference tasks; this would be the first large-scale training adoption.
- Large-scale validation on the inference side has boosted confidence in domestic chips.
- Ecosystem migration costs on the training side remain the key variable limiting the market ceiling.