DeepSeek V5 reportedly to be pretrained on Huawei Ascend chips, Ascend 960DT due Q1 2027
teortaxesTex · x · 2026-11-03
According to multiple industry sources cited by Science and Technology Innovation Board Daily, DeepSeek's next-gen large model V5 is planned to be pretrained on Huawei Ascend chips, which previously were used mainly for inference in DeepSeek models.
The leaker notes Ascend 960DT is expected to be ready in Q1 2027; a commenter adds that an Nvidia cluster large enough for DeepSeek's plans is no longer feasible, so the rumored 8T-parameter model will be trained on Huawei NPUs.
Related event: DeepSeek V5 Reportedly to Use Huawei Ascend Chips for Pretraining(2 posts)→
More from Infra
- FLAMIA: MXFP8 microscaling attention kernel for fast, causally-safe training on Blackwell — yresearch · 2026-11-03
- 2,500+ GW of power projects stuck in grid queues as $200B flows into energy tech — ingliguori · 2026-11-03
- Build a Media Agent workshop: CPU-based agent calling GPU tools on demand — Al_Grigor · 2026-11-03
- Factory rebuilt its Droid coding agent to run inside customers' own clouds, plans to open source the compute layer — AI Engineer · 2026-11-03
- Korean HBM smuggled into China: The Information investigates — AccBalanced · 2026-11-03
- Researchers dissect NVIDIA GPU confidential computing and test its security guarantees — burkov · 2026-11-03