DeepSeek has released tools for Huawei chips that could replace Nvidia's solutions

The Chinese AI startup DeepSeek has released open-source software for programming AI chips, developed in collaboration with Huawei, according to Bloomberg.

The software toolkit, which is now available for free download, is designed for Huawei’s Ascend accelerators. The toolkit includes TileLang—the Chinese equivalent of Nvidia’s CUDA platform, which is considered the global standard for developing artificial intelligence software.

The release highlights the growing collaboration between two leading Chinese AI companies seeking to replace American software and hardware solutions with domestic alternatives, Bloomberg notes. DeepSeek is developing open-source models that compete with those from Anthropic and OpenAI, while Huawei is leading China’s efforts to manufacture accelerators for data centers.

A complex programming language with a simple coding process is essential for building an effective AI ecosystem, DeepSeek stated. The startup has already worked with Huawei to optimize performance on Ascend 950 chips. Huawei provided “unconditional and vigorous support” during the software’s development, and the two companies intend to continue collaborating on new solutions, DeepSeek added. The company itself rarely discloses details of its internal development, including the integration of AI chips.

Although global model training still relies heavily on Nvidia processors, DeepSeek is actively transitioning to domestically produced hardware. The startup plans to deploy at least 160,000 of Huawei’s most powerful accelerators in a large data center it is building in Inner Mongolia, Bloomberg reported last month. At the same time, the company is finalizing a funding round at a valuation of about 500 billion yuan ($74 billion) ahead of a possible initial public offering as early as this year.

This article was AI-translated and verified by a human editor