DeepSeek has publicly released software it created with Huawei Technologies Co to program artificial intelligence (AI) chips, showcasing its close partnership with the Chinese company working to develop technologies to replace Nvidia Corp.
The AI lab is making its software toolkit for Huawei’s Ascend accelerators open-source and free to download, the startup said on its WeChat account on Sept 30. Those tools include TileLang, China’s answer to Nvidia’s CUDA: a software platform considered the global standard for developing AI.
The move reflects deepening ties between two of China’s leading AI firms, among a handful of players trying to fulfil Beijing’s vision of replacing American software and circuitry with local equivalents. While DeepSeek is developing open-weight models to compete with the likes of Anthropic and OpenAI, Huawei is leading China’s effort to produce the accelerators essential to data centres.
DeepSeek has actively adopted domestic chips, even though model training worldwide still relies heavily on semiconductors from Nvidia. The Chinese startup plans to deploy at least 160,000 of Huawei’s most powerful accelerators in a massive data centre it’s building in Inner Mongolia, Bloomberg News reported last month. It’s also finalising a round of fundraising at a valuation close to 500bil yuan (US$74bil or RM302bil), ahead of a public listing as early as this year.
DeepSeek rarely talks about internal development, including AI chip integration.
A sophisticated programming language with a simple coding process is essential for building an effective AI ecosystem, DeepSeek said in its WeChat post. The company has worked with Huawei to optimise performance on Ascend 950 chips, the startup added.
Huawei provided "unreserved and vigorous support” during the research and development of the software, and the two will collaborate on more innovations, DeepSeek said. – Bloomberg
