Noesion Radarnoesion.ai
Research2026-11-03

DeepSeek plans to pre-train its next model on Huawei Ascend chips. This would be the first time Ascend chips are used for pre-training a DeepSeek model. Huawei deployed 3,000 Ascend chips to support DeepSeek V3 and R1 in early 2025. Huawei later said its systems could fully support DeepSeek V4. DeepSeek open-sourced tools adapted for Huawei’s Ascend platform in September. A Shenzhen consortium completed post-training of DeepSeek V4-Pro on Ascend 910C chips. Pre-training from scratch remains untested on Ascend chips. Success would reduce reliance on Nvidia hardware for Chinese AI developers. The plan has not been independently confirmed. Viability depends on training stability, cluster performance and software maturity.

What changed

Ascend chips were used mainly for inference and post-training, not pre-training.

What it unlocks

Pre-training a next-generation model on domestic Chinese AI chips.

Sources

Send this to someone who needs it

Shares the story and its sources. Nothing about you.

What does this mean for your job?

This is the story as everyone gets it. Once a week we send you the version written for your role — what changed, why it matters for the work you actually do, and one thing to try. Free while we tune it.