Chinese AI chip stocks rise as DeepSeek releases software for domestic chips
Cambricon rose more than 8% as DeepSeek's open-source DeepInfer framework, which runs on Huawei Ascend chips, lifted China's chip-design stocks.
The short version
- Cambricon led gains of more than 8% on November 9; the STAR chip design index rose more than 2%.
- DeepInfer is DeepSeek's open-source framework for running models in large cloud services.
- It supports distributed inference, speculative decoding and structured output.
- It runs on NVIDIA and Huawei Ascend platforms; DeepSeek also released a token encoding and decoding component.
- The outlet's own explanation is promotional in tone, and a market move is not proof of performance.
Why it matters
AI chips are only useful if the software that runs models on them works well. When a leading Chinese lab releases open tools that run on domestic hardware, it narrows the software gap that has favoured NVIDIA. Treat one day's share move as a mood signal, not as evidence that the chips got faster.
What moved on the market
As Jinrongjie (Financial Circles) reported, Chinese computing-chip stocks strengthened during trading on November 9. The STAR Market chip design index, which tracks chip-design companies listed on Shanghai's tech board, rose more than 2%.
Cambricon, a Chinese AI chip maker, led with a gain of more than 8%. Fudan Micro and MetaX also gained.
What DeepSeek released
The outlet points to DeepInfer, an open-source framework from DeepSeek released earlier. It is built for inference, meaning running an already-trained model to answer requests, at cloud-service scale.
It supports spreading the work across many machines. It also supports speculative decoding, a speed-up where the system drafts likely next words and then checks them. A third feature is structured output, where the model replies in a fixed format. DeepSeek released a text-to-token conversion component the same day.
DeepInfer runs on NVIDIA hardware and on Huawei's Ascend platform.
Why Ascend support matters
The report argues that open software for Ascend fills a gap in the tools around domestic chips. Companies could find it easier to build clusters of Ascend machines for serving models.
The speed-up features are said to raise throughput and efficiency, which lowers running costs. The outlet also says that open code encourages developers to adopt both the models and the chips.
This is the article's own reading, and it comes from a piece that also promotes a chip-design fund. The report gives no benchmarks for how DeepInfer performs on Ascend.
How to read the price move
The report links the rally to the framework, but it does not show that one caused the other. Intraday figures also change by the hour.
The article carries standard fund risk warnings and says nothing in it is investment advice. The same caution applies here.
What to watch
- Whether independent tests show how DeepInfer performs on Ascend compared with NVIDIA chips.
- Whether other Chinese chip makers announce support for DeepSeek's open tools.
Try it
When you read about a model or chip, check which hardware its software actually runs on, since that decides whether you can use it.
Sources and links
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by Jinrongjie (Financial Circles) · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.