Huawei has reportedly developed a new technological solution for AI inference, aiming to mitigate China’s reliance on high-bandwidth memory (HBM) chips. The company plans to unveil this innovation at the 2025 Financial AI Reasoning Application Forum on August 12.
The “Financial AI Reasoning Application Implementation and Development” event serves as a platform for technology vendors to demonstrate their AI capabilities and present practical applications of AI within the financial sector. Huawei’s forthcoming announcement aligns with this focus, introducing a solution designed to address current dependencies on foreign technology.
AI inference is the operational phase of an AI model, wherein it applies its acquired knowledge and learned methodologies to generate precise and efficient outputs for end-users. High-bandwidth memory chips are typically favored for AI inference due to their reduced latency and increased memory bandwidth compared to conventional memory types. This characteristic facilitates expedited data processing, thereby enhancing the performance of large language models and improving the overall AI inference process. However, Huawei’s access to HBM chips has been restricted following US sanctions.
In response to these restrictions, Huawei has reportedly engineered a technological solution in the domain of AI inference. This innovation is anticipated to decrease China’s and Huawei’s reliance on HBM AI chips while simultaneously boosting the inference performance of large-scale AI models within the country. This development is expected to significantly strengthen China’s domestic AI inference ecosystem.
Huawei continues to pursue self-developed technological integrations and solutions to diminish its reliance on US-origin components as it expands its AI business within China. Details regarding the new AI inference technology currently remain undisclosed, with further information expected at the upcoming August 12 event.
