Huawei is reportedly set to unveil a new technological solution aimed at reducing China’s reliance on High Bandwidth Memory (HBM) chips for Artificial Intelligence (AI) inference. This innovation will be presented at the 2025 Financial AI Reasoning Application Forum on August 12.
AI inference, the “doing” part of AI, involves models using their knowledge to efficiently deliver accurate outputs. HBM chips are crucial for this process due to their lower latency and higher memory bandwidth compared to traditional memory, which facilitates faster data processing and improved performance for large language models.
However, Huawei has faced limitations in accessing HBM chips due to US restrictions. In response, the company has reportedly developed a proprietary solution designed to circumvent this dependency. This new technology is expected to not only lessen China’s and Huawei’s reliance on imported HBM AI chips but also significantly boost the inference performance of large-scale AI models within the country.
The development is seen as a strategic move to strengthen China’s domestic AI inference ecosystem. Huawei has been actively seeking self-developed technological integrations to expand its AI business within China, aiming to reduce its reliance on foreign goods. While the specific details of the new solution remain undisclosed, further information is anticipated to be revealed at the upcoming forum.
