DeepSeek Open-Sources Infrastructure Components for Huawei Ascend Computing Platform, Corresponding One-to-One with Nvidia Platform
DeepSeek has officially open-sourced infrastructure components for Huawei's Ascend computing platform, covering tools such as TileLang. These components correspond one-to-one with Nvidia platform components and have already supported the training of the DeepSeek V4 model.
Key Takeaways
- Key Highlight：DeepSeek has officially open-sourced infrastructure components for Huawei's Ascend computing platform, covering tools such as TileLang. These components correspond one-to-one with Nvidia platform components and have already supported the training of the DeepSeek V4 model.
- Innovation & Tech：Highlights advancements in DeepSeek, Open-Sources, Infrastructure, demonstrating rapid progress in model capabilities.
- Industry Impact：Reported via IT之家 (CN), offering actionable signals for developers and technology leaders.
DeepSeek announced the open-sourcing of infrastructure components adapted for Huawei's Ascend computing platform, including the TileLang high-level language compilation tool, compute libraries, and distributed communication libraries. These components correspond to previously open-sourced components for the Nvidia platform, aiming to provide complete underlying software support for the Ascend platform.
As a core tool, TileLang benchmarks against Nvidia's CUDA but optimizes programming complexity to improve development efficiency. This programming model can fully leverage the underlying characteristics of the chip, approaching hardware performance limits. It has already been validated on the Nvidia platform and applied to the training of the DeepSeek V4 series models.
This move marks progress for domestic computing platforms in building AI infrastructure ecosystems. By transferring its mature experience from the Nvidia platform to the Ascend platform, DeepSeek helps lower the technical threshold for developers to deploy and train large models on domestic chips.
As leading large model enterprises deeply participate in the software optimization of domestic computing infrastructure, the software ecosystem shortcomings of domestic AI chips are expected to be bridged more rapidly. This not only provides more computing power options for domestic large model training but will also promote the practical application and iteration of domestic computing power in high-end AI training scenarios.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.
Industry Insights & Analysis
As artificial intelligence rapidly evolves, breakthroughs surrounding DeepSeek, Open-Sources, Infrastructure, Components are shifting toward scalable, robust real-world implementations.
Driven by both open-source ecosystems and proprietary model architectures, the integration between compute optimization, data engineering, and agentic workflows is accelerating. This development provides a strategic benchmark for upcoming AI tooling and developer workflows.