Skip to content

Huawei set to reveal technology aimed at reducing China's dependence on HBM chips, according to a recent report.

Although China's leading memory chip producers are advancing, they continue to trail behind American and Korean counterparts like Micron and SK Hynix in the competition.

Huawei set to reveal technology reducing China's dependence on HBM chips, according to a report
Huawei set to reveal technology reducing China's dependence on HBM chips, according to a report

Huawei set to reveal technology aimed at reducing China's dependence on HBM chips, according to a recent report.

Huawei and China UnionPay Collaborate to Advance AI in Financial Sector

Huawei Technologies is set to make an announcement at the 2025 Financial AI Reasoning Application Landing and Development Forum in Shanghai, in collaboration with China UnionPay. This collaboration aims to promote AI reasoning models and applications in the financial sector.

At the heart of this collaboration is a new software tool called the Unified Cache Manager (UCM), designed to accelerate AI inference without relying heavily on high-bandwidth memory (HBM) chips. This tool optimizes AI data management across different memory types—HBM, standard DRAM, and SSD—based on application latency needs.

The UCM has been trialed in real operational scenarios at China UnionPay, with significant improvements reported. These include up to 90% latency reduction and a 22x throughput increase for AI inference on traditional memory architectures. This practical step towards software-driven mitigation of hardware constraints could potentially alleviate the impact of US export restrictions on cutting-edge HBM chips.

The partnership's focus on financial AI applications such as fraud detection and algorithmic trading highlights an immediate use case for these innovations in China's financial technology ecosystem. Huawei's wider strategy is linked to responding to the ongoing US sanctions that restrict access to advanced AI chip technology produced by US and South Korean firms.

If the development at the forum represents a step by Huawei Technologies to establish a self-sufficient AI hardware ecosystem in China, it may involve the use of HBM semiconductors supplied by Yangtze Memory Technologies and Changxin Memory Technologies. These Chinese companies are potential contributors to Huawei's efforts to reduce reliance on US and South Korean suppliers for HBM semiconductors.

The forum is scheduled to take place on Tuesday. Huawei did not immediately respond to a request for comment. The top suppliers of HBM semiconductors, apart from Yangtze Memory Technologies and Changxin Memory Technologies, are US companies Micron Technology and Advanced Micro Devices, as well as South Korean firm Samsung Electronics and SK Hynix.

This collaboration between Huawei and China UnionPay marks a significant move towards technological self-sufficiency in AI hardware and reasoning models within the Chinese financial sector. This development potentially diminishes the critical dependence on foreign-made HBM memory suppliers, especially those based in the US and South Korea.

  1. The new software tool developed by Huawei, the Unified Cache Manager (UCM), is not only intended for AI inference without relying on high-bandwidth memory (HBM) chips, but also for optimizing AI data management across various memory types, including artificial-intelligence-related applications.
  2. In the arts sector, this technological advancement may inspire artists and designers to contemplate the integration of AI-enhanced tools in producing innovative and thought-provoking creations, further blurring the lines between humanity and artificial intelligence.
  3. As Huawei's partnership with China UnionPay progresses, financial businesses worldwide might take notice and consider similar collaborations to harness the power of AI in their operations, potentially creating a ripple effect that transforms the global financial landscape.

Read also:

    Latest