DeepSeek Partners With Huawei to Develop Homegrown AI Training Chips

A Chip Collaboration Takes Shape

DeepSeek is working with Huawei Technologies to design AI chips tailored to its models, a strategic push that could help China's leading AI startup train advanced systems without relying on Nvidia, The Information reported. Huawei engineers have been stationed at DeepSeek's offices, incorporating its feedback into chip designs, including how AI models perform calculations and transfer data between processors, according to people familiar with the collaboration.

DeepSeek plans to begin using Huawei's Ascend 950DT chips for model training as soon as the end of 2026. Huawei also plans to prioritize DeepSeek for most of its next-generation Ascend 960DT chips, expected in the first quarter of 2027. The partnership supports Beijing's push to build an AI industry independent of American technology, even as Nvidia remains central to training China's most advanced models despite U.S. export restrictions.

Technical Gap Remains Wide

The technical hurdles remain significant. Nvidia's B300 chips deliver roughly seven times the raw arithmetic performance of Huawei's Ascend 950 processors, according to research institute Epoch AI. Training a frontier model could require more than 10,000 Huawei chips operating together for weeks, making manufacturing capacity and system reliability critical challenges for the partnership.

Reducing Reliance on CUDA

Beyond hardware, DeepSeek is also using TileLang, an open-source programming language that can help developers adapt AI software across chip architectures, reducing reliance on Nvidia's CUDA ecosystem. Last month, it released a version compatible with Huawei's Ascend 950 chips.

DeepSeek is not abandoning Nvidia and continues to seek access to its advanced processors, The Information reported. But if the collaboration enables DeepSeek to train a frontier model entirely on Chinese chips, it could offer a blueprint for other AI developers and mark a major step toward reducing China's dependence on U.S. technology.

Sources: Yahoo Finance · The Information