DeepSeek R1 Is Reportedly Running Inference On Huawei’s Ascend 910C AI Chips, Showing China’s Growing AI Capabilities

Jan 28, 2025 at 11:34am EST

DeepSeek's AI model reportedly runs inference workloads on Huawei's newest Ascend 910C chips, showing how China's AI industry has evolved over the past few months.

DeepSeek's AI Compute Portfolio Includes Huawei's Ascend AI Chips As Well, Running Inferencing Workloads Through It

Well, the Chinese AI firm DeepSeek has surely managed to disrupt the global AI markets over the past few days, as their recently-announced R1 LLM model managed to shave off $2 trillion from the US stock market since it created a sense of panic among investors. While claims around the compute power DeepSeek used to train their R1 model are pretty controversial, it seems like Huawei has played a big part in it, as according to @dorialexander, DeepSeek R1 is running inference on the Ascend 910C chips, adding a new twist to the fiasco.

Related Story Tenstorrent Vows to ‘Crush Everyone’ as Galaxy Blackhole Hits 350 Tokens/s on DeepSeek R1, Undercutting NVIDIA’s GB300 5x AI TCO

For those unaware, Huawei's Ascend 910C AI chip is said to be a direct rival to NVIDIA's Hopper H100 AI accelerators, and while the specifics of Huawei's chip aren't certain for now, it was claimed that the company planned to start mass production in Q1 2025, seeing interest from mainstream Chinese AI firms like ByteDance and Tencent. Huawei's AI chips are known to be the top-tier alternative to NVIDIA's hardware in China, and they have managed to gobble up a hefty market share, so it seems like they will become a lot more popular.

The computing resources used around DeepSeek's R1 AI model are not specific for now, and there's a lot of misconception in the media around it. Firstly, the "$5 million" figure isn't the total training cost but rather the expense of running the final model, and secondly, it is claimed that DeepSeek has access to more than 50,000 of NVIDIA's H100s, which implies that the firm did require resources similar to other counterpart AI models.

Utilizing Huawei's chips for inferencing is still interesting since not only are they available in ample quantities to domestic firms, but the pricing is pretty decent compared to NVIDIA's "cut-down" variants or even the accelerators available through illegal sources. Huawei is said to be developing the next generation of Ascend AI chips, which are said to rival Team Green's Blackwell AI products and will undoubtedly ramp up global competition.

About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.

Follow Wccftech on Google to get more of our news coverage in your feeds.