TOKYO -- Kioxia Corporation announced that its KIOXIA AiSAQ™, which is Approximate Nearest Neighbor Search (ANNS) software technology has been integrated into the open-source vector database Milvus beginning with version 2.6.4. With this integration, Milvus users can take full advantage of KIOXIA AiSAQ™ SSD-optimized vector search capabilities, providing developers and enterprises with a practical, cost-efficient path to scaling AI applications without facing the difficulty of scaling DRAM memory size typically associated with large-scale vector search.

The AI industry is shifting from building massive foundation models to deploying scalable, cost-effective inference solutions that address real-world challenges. RAG (Retrieval Augmented Generation) is central to this transition, and KIOXIA AiSAQ™ technology was developed to help the community leverage SSD-based vector architectures. Its integration into the Milvus ecosystem enhances ease of adoption within the open-source community and supports developers building faster, more efficient AI applications.

First announced earlier this year, KIOXIA AiSAQ™ is an open-source software technology designed to dramatically improve vector scalability by storing all RAG-related database elements on SSDs*[1]. As DRAM scalability has become a critical bottleneck for high-volume inference and RAG workloads, KIOXIA AiSAQ™ technology provides a breakthrough by sharply reducing DRAM requirements while maintaining high-quality vector search accuracy.

With KIOXIA AiSAQ™ technology now integrated into Milvus, Kioxia and the open-source community are enabling a new class of scalable, cost-efficient vector search solutions designed to meet the rapidly growing demands of modern AI applications.