KIOXIA Scales Vector Search to 4.8 Billion Vectors on Single Server, Cuts Index Build Time by 7.8x — KIOXIA America, Inc.

Event summary

KIOXIA demonstrated high-dimensional vector search scaling to 4.8 billion vectors on a single server using its open-source KIOXIA AiSAQ™ technology.
GPU acceleration via NVIDIA cuVS reduced index build time by up to 7.8x, cutting end-to-end build times from 31 days to 4 days.
The achievement addresses scalability challenges in retrieval augmented generation (RAG) search solutions.
KIOXIA AiSAQ technology enables vector search directly from SSDs, reducing DRAM usage and supporting large-scale deployments.
The company is continuing development to support deployments beyond 4.8 billion vectors.

The big picture

KIOXIA's breakthrough in vector search scalability addresses a critical bottleneck in AI applications requiring real-time processing of massive unstructured datasets. By leveraging SSD-based storage and GPU acceleration, the company is positioning itself as a key player in the evolving AI infrastructure landscape. This development comes as demand for high-dimensional vector databases grows, particularly in retrieval augmented generation (RAG) systems.

What we're watching

Scalability Limits: Whether KIOXIA can sustain further improvements in vector search capacity beyond 4.8 billion vectors.
GPU Dependency: How reliance on NVIDIA GPU acceleration may impact KIOXIA's flexibility and cost structure.
Market Adoption: The pace at which enterprises adopt KIOXIA's SSD-based vector search for large-scale AI applications.