DeepSeek Open-Sources AI Tools for Huawei Ascend Chips

DeepSeek Releases Six Open-Source Tools for Huawei Ascend AI Chips, Challenging Nvidia 

HANGZHOU, China September 30, 2026, Chinese artificial intelligence start-up DeepSeek on Wednesday open-sourced a suite of six software tools tailored for Huawei Technologies' Ascend AI chips, marking a pivotal step in China's effort to reduce dependence on processors from US semiconductor giant Nvidia amid ongoing export restrictions. The development carries direct implications for the global Semiconductor Market, valued at USD 639.10 billion in 2023 and projected to reach USD 1,209.35 billion by 2030, at a CAGR of 9.57%. 

Hangzhou-based DeepSeek released the tools with the stated aim of building a new "independent and controllable" software ecosystem for graphics processing units (GPUs), according to a post on its official WeChat account. Among the key releases is an Ascend-compatible version of TileLang a custom programming language designed to streamline the development of high-performance computation kernels which now officially supports Huawei's Ascend 950 accelerators, offering "native code generation, automatic scheduling, and synchronisation." 

Additional tools released include DeepGEMM Ascend, a matrix multiplication kernel library, and DeepEP Ascend, which enables efficient, large-scale cross-device communication. DeepSeek stated that the new tools delivered computational and communication performance approaching "the hardware limits" in several benchmark tests, underscoring the technical maturity of the release. 

The development was made possible through "close collaboration" with Huawei, with the two companies jointly optimizing a supernode architecture powered by 128 Ascend 950 processors. Huawei has separately announced that its Ascend 960DT chip designed for AI model training will be "ready in the first quarter of 2027," three quarters ahead of its original schedule, further reinforcing the momentum behind China's domestic AI chip ecosystem. 

Key Highlights: 

• DeepSeek released six open-source software modules mirroring its prior tools for Nvidia's AI chips, now fully adapted for Huawei's Ascend 950 accelerators. 

• TileLang's Ascend version provides "native code generation, automatic scheduling, and synchronisation," enabling high-performance AI workloads on domestic hardware. 

• The initiative reflects Beijing's broader semiconductor self-reliance drive, with Chinese chip firms including Cambricon Technologies, Moore Threads, and Alibaba's T-Head all confirming compatibility with DeepSeek models. 

• Huawei's Ascend 960DT AI training chip is now scheduled for Q1 2027 delivery three quarters ahead of the original timeline accelerating the domestic AI hardware roadmap. 

Analyst Insight: 

According to analysts at Next Move Strategy Consulting, DeepSeek's open-sourcing of Huawei-compatible AI software tools represents a structurally significant development for the global semiconductor landscape. By establishing a parallel software ecosystem for non-Nvidia AI processors, Chinese technology firms are accelerating the bifurcation of the global chip supply chain a trend that is likely to intensify competitive dynamics and reshape procurement strategies across data center operators and AI infrastructure providers worldwide. 

NMSC analysts note that as AI-driven demand continues to expand, the ability to deploy high-performance software stacks on domestically produced hardware will become an increasingly critical differentiator in markets subject to export controls. This development may also prompt accelerated investment in software-hardware co-optimization strategies among semiconductor players globally, as the competitive moat around proprietary software ecosystems long a key advantage for Nvidia  faces a more credible challenge. 

Industry Outlook: 

The open-sourcing of Huawei-compatible AI tools by DeepSeek signals a maturing of China's domestic semiconductor software ecosystem, which has historically lagged behind hardware development. As Chinese AI firms deepen collaboration with domestic chip manufacturers, competitive pressure on established global semiconductor players is expected to intensify across both hardware and software dimensions. 

For the broader semiconductor industry, this development underscores the growing strategic importance of software-hardware co-optimization particularly in markets where hardware access is constrained by geopolitical factors. The trajectory of China's semiconductor self-sufficiency drive will remain a key variable shaping global chip market dynamics through the remainder of the decade, with implications for supply chain diversification, technology licensing, and cross-border investment flows. 

Source: South China Morning Post 

For More Information: Download FREE Sample on Semiconductor Market Report

Prepared By: Sanyukta Deb