Chinese AI firm DeepSeek has partnered with Huawei Technologies to develop programming tools for Huawei's Ascend chips. This collaboration aims to create an independent GPU software ecosystem and reduce dependence on Nvidia's CUDA. DeepSeek is also developing its own AI inference chip.
The partnership between DeepSeek and Huawei, along with DeepSeek's development of its own AI inference chip, signifies a broader trend of Chinese AI companies seeking technological independence and mitigating risks associated with geopolitical tensions and export controls, potentially reshaping the global AI hardware and software landscape.
Chinese AI firm DeepSeek announced a partnership with Huawei Technologies to develop programming tools for Huawei's Ascend AI chips, signaling a move to build an independent software ecosystem and lessen reliance on Nvidia. The collaboration involves open-sourcing DeepSeek's TileLang programming language and related libraries for the Ascend platform. DeepSeek stated that TileLang is designed to offer a simpler programming model than Nvidia's CUDA, aiming to enhance development efficiency and code logic.
This initiative comes as Chinese tech companies deepen ties to find alternatives to the Nvidia ecosystem, particularly in light of U.S. export controls. Sources cited by Reuters indicate that DeepSeek began developing its own AI inference chip around mid-2025, a strategic shift to gain hardware self-sufficiency. This in-house chip development effort is intended to reduce structural dependence on both Nvidia and Huawei. The company has been actively engaging with chip design firms, foundries, and memory suppliers, and has been discreetly hiring chip design engineers. DeepSeek's founder, Liang Wenfeng, has previously highlighted chip access as a significant constraint, underscoring the importance of controlling the supply chain.
Pick the topics you care about. Get only what matters, on your cadence.