This system promotes advanced automation and efficiency in logistics, enabling organizations to respond quickly to evolving market needs while managing costs and maintaining service quality. KIOXIA AiSAQ™ [1] and memory-centric AI [2] technologies are central to this endeavor, addressing the need for scalable AI as product types continue to expand and diversify. The jointly developed technology will be showcased at the 2025 International Robotics Exhibition. With the continued growth of e-commerce transactions, logistics networks are experiencing increased volume and a wider variety of products. Simultaneously, ongoing labor shortages are driving the need for greater operational efficiency through AI. Traditional image recognition AI systems rely on deep learning models that require parameter adjustments and retraining each time new or seasonal products are introduced. This process is time-consuming and increases both energy consumption and operating costs, especially when dealing with large product catalogs. KIOXIA AiSAQ software , combined with KIOXIA's memory-centric AI technology, addresses these challenges by storing large amounts of new product data, including images, labels, and feature information, in high-capacity storage. This allows for the rapid addition of new product information without requiring retraining of the base model. To mitigate longer lookup times and increased memory requirements as data volume grows, the technology indexes the data stored in memory and moves it to SSD storage, enabling faster and more efficient retrieval. “At KIOXIA, our goal is not only to provide the best memory options for application needs, but also to offer support and accessibility through the open source of our technology, in order to help developers and system architects fine-tune performance and capacity in new and innovative ways,” stated Axel Störmann, Vice President and Head of Product Technology for Memory and SSDs at KIOXIA Europe GmbH. “By using SSD-based ANNS, we are reducing our reliance on expensive DRAM while simultaneously meeting the performance requirements of leading memory solutions, significantly improving the performance range of large-scale RAG applications.”
[1]: KIOXIA AiSAQ™ technology, designed to reduce DRAM requirements in generative AI systems, is released as open source software
https://www.kioxia.com/en-jp/business/news/2025/20250128-1.html
[2]: Development of an image classification system using memory-centric AI with high-capacity storage
https://www.kioxia.com/en-jp/rd/technology/topics/topics-39.html
