DeepSeek V5 plans to conduct pre-training based on Huawei Ascend chips

Cailian Press Financial News Agency

Open APP

Domestic chips training domestic models! DeepSeek's next-generation large model is planned to be trained on Ascend chips

STAR Market Daily reporter Huang Xinyi, editor Mao Mingjiang, 2026-11-03 14:24

① Learned from multiple industry chain sources that DeepSeek's next-generation large model DeepSeek V5 is planned to be trained based on Ascend chips. Previously, Ascend chips were mainly used for inference tasks of DeepSeek large models. ② The large-scale validation of domestic chips on the inference side has provided industry confidence, but the ecosystem migration cost on the training side remains the variable that determines the market ceiling.

STAR Market Daily, November 3 (Reporter Huang Xinyi) — The STAR Market Daily reporter learned from multiple industry chain sources that DeepSeek's next-generation large model DeepSeek V5 is planned to be pre-trained based on Ascend chips. Previously, Huawei's Ascend chips were mainly used for inference tasks of DeepSeek large models.

Judging from publicly available information, the earliest large-scale deployment of Ascend carrying DeepSeek models occurred around the Spring Festival of 2025.

Fang Xingdong, executive vice dean of the College of Media and International Culture at Zhejiang University, mentioned in the book "The Rise of Ascend" that on January 28, 2025, Huawei's senior management issued a decision for Huawei Cloud to quickly support DeepSeek. From New Year's Eve to the second day of the Lunar New Year, Huawei and DeepSeek completed the deployment and validation of more than 3,000 Ascend chips, successfully deploying the DeepSeek-V3/R1 large models to Huawei Cloud's Gui'an Data Center. The book also writes that DeepSeek's explosive growth greatly excited Ren Zhengfei, and throughout the Spring Festival he was directing Huawei Cloud and related product teams to cope with the traffic tsunami. In January of the same year, SiliconFlow and Huawei Cloud launched DeepSeek-V3 and R1 inference services based on Ascend.

After DeepSeek-V4 was released in April 2026, Huawei stated that through the two sides' "chip-model technology collaboration," the full series of Ascend supernode products could fully support the DeepSeek V4 series models. DeepSeek mentioned on its official API page that, limited by high-end computing power, the service throughput of the V4-Pro model is still limited at present, and it is expected that after the Ascend 950 supernodes are mass-produced and launched in the second half of the year, the Pro price will be significantly reduced.

On September 30, DeepSeek announced the official open-sourcing of infrastructure components for the Huawei Ascend computing platform, covering the TileLang high-level language compilation tool, high-performance computing libraries (DeepGEMM, FlashMLA, TileKernel, DeepSelect, etc.) and the distributed communication library (DeepEP), corresponding one-to-one with the previously open-sourced components for the NVIDIA platform. Huawei provided the jointly defined Ascend supernode SuperPoD Flex and UBL128 networking solutions, and opened the Ascend C programming interface and underlying instruction system to support TileLang's work.

The significance of this move is that DeepSeek is not only using Ascend for inference, but has also systematically aligned and co-built with Ascend at the underlying software stack level, extending adaptation to the infrastructure level that supports training and inference.

At present, the implementation of domestic computing power is mainly concentrated on the inference side, and still needs improvement on the training side. Not long ago, the AI training platform project team of Shenzhen Hetao College, together with relevant teams from Harbin Institute of Technology (Shenzhen), the Shenzhen Big Data Research Institute, and Huawei, collaborated with the Shenzhen Smart City AI computing platform and, relying on the Ascend 910C domestic AI computing cluster, completed full-parameter post-training of the 1.6-trillion-parameter large model DeepSeek-V4-Pro. It should be noted that this training still did not involve basic pre-training.

The large-scale validation of domestic chips on the inference side has provided industry confidence, but the ecosystem migration cost on the training side remains the variable that determines the market ceiling. If DeepSeek's next-generation large model completes pre-training based on Huawei Ascend computing power, it will for the first time answer the question "whether China's top large models can complete pre-training tasks on domestic computing power," a question that previously had only doubts and no answers.

Cailian Press statement: The content of this article is for reference only and does not constitute investment advice. Investors who act on it do so at their own risk.

Business cooperation

Popular unlocks

[Intraday Treasure] As policy planning details gradually land, institutions say order certainty in this industry may increase, and multiple sub-sectors are expected to usher in high-prosperity development. This enterprise has the capability to develop and produce multiple types of products.

As policy planning details gradually land, institutions say order certainty in this industry may increase, and multiple sub-sectors are expected to usher in high-prosperity development. This enterprise has the capability to develop and produce multiple types of products, while another enterprise has completed delivery of sub-sector products, with business scale reaching a record high.

[Intraday Treasure] Bookings are earlier and growth is faster this year, and this sector continues to heat up as the peak season arrives. This company has relevant resource advantages.

Customers are entering early, booking volume has doubled year on year, and this industry continues to heat up as the peak season arrives. This company has relevant resource advantages.

[Trend Research Report · Company] Fresh snack formats such as Jinlimen are hot, and this long-established chain leader is actively polishing its fresh snack store model. Feedback from traditional store adjustments is good, and next year it is expected to see a "turnaround" on the operating side.

Fresh snack formats represented by Jinlimen are hot, and this long-established food chain leader is actively polishing its fresh snack store model. Feedback from traditional store adjustments is good and gradually being rolled out, and next year it is expected to see a "turnaround" on the operating side. Analysts are initiating coverage.

① According to Cailian Press Chuangtoutong data, before 2025, primary-market financing for physical AI/world models was mostly in the single digits, rising to 23 deals in 2025 and reaching more than 100 deals in 2026. ② Multiple investors said they have indeed become more cautious, with investment focus shifting from the imaginative space of technology to investment risk and business delivery capability. But the slowdown in pace still has a certain lag before it is reflected in the financing data on paper.

Terrified by the tech stock rout! Wall Street strongly advises Anthropic to price prudently: retreat to $1.5 trillion

Cailian Press, Xiaoxiang, 2026-11-03 09:20

① According to people familiar with the matter, in the past few days, institutional investors have indicated to this Claude developer and its advisory team that in the IPO later this month, its valuation should be set at around $1.5 trillion to $1.6 trillion. ② Investors cited multiple concerns, including intensified market turbulence and growing worries about excessive concentration of investment in the AI sector.

Cailian Press, Huang Junzhi, 2026-11-03 09:23

① Hassett acknowledged that some companies may be carrying huge debts, and this phenomenon is indeed worth watching; ② He also pointed out that there are many financially strong companies in the industry that hold both equity and debt in the relevant companies. If a company gets into trouble, these companies can step in to acquire it or provide funding support.