AI BriefingDeepSeek Plans to Pre-Train Next-Gen Model DeepSeek V5 on Huawei Ascend Chips
·2026.11.03 19:00
Key point
The move would mark the first time Ascend chips are used for full pre-training, following a September 2026 release of open-source infrastructure components adapted for the platform.
Details
- DeepSeek plans to use Huawei Ascend AI chips to pre-train its next-generation model, DeepSeek V5, a shift from previous use of Ascend primarily for inference.
- The plan is based on a report by STAR Market Daily citing industry sources and has not yet been independently confirmed.
- In September 2026, DeepSeek open-sourced infrastructure components for the Ascend platform, including compiler tools, high-performance computing libraries, and distributed communication software.
- A Shenzhen research consortium recently completed full-parameter post-training of DeepSeek V4-Pro, a 1.6-trillion-parameter model, on a cluster of Huawei Ascend 910C chips.
- Huawei deployed more than 3,000 Ascend chips to support DeepSeek V3 and R1 in early 2025.
- Successful pre-training would test whether Chinese AI accelerators can support frontier LLM pre-training, potentially reducing dependence on Nvidia hardware.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
DeepSeek V4.1 Flash Leads Agentic Coding Benchmarks, Narrowing US-China AI Gap to 3%DeepSeek Nears $12B Funding Round With Tencent and CATL Among Backers