Kioxia Achieves 4.8 Billion Vector Search

Kioxia announced it demonstrated high-dimensional vector search scaling to 4.8 billion vectors on a single server using its open-source KIOXIA AiSAQ combined with NVIDIA cuVS, reporting up to 20x faster index build time for 1024-d vectors using four Hopper GPUs (index build reduced from 28.4 to 1.4 days). Kioxia says end-to-end build times improved up to 7.8x, enabling SSD-based billion-scale RAG deployments with minimal DRAM.