Staff Customer Qualification Engineer

KIOXIA achieves 4.8 billion high-dimensional vector search database on a single server, with 7.8x index build time acceleration via GPUs. Kioxia Corporation Leveraging the NVIDIA cuVS Library with KIOXIA AiSAQ Technology to Index Vectors of 1024 Dimensions with Minimal DRAM Use TOKYO-BUSINESS WIRE- Kioxia Corporation today announced the successful demonstration of achieving high-dimensional vector search scaling to 4.8 billion vectors on a single server with its open-source KIOXIA AiSAQ(TM) approximate nearest neighbor search (ANNS) technology. Additionally, Kioxia demonstrated a significant reduction in index build time by leveraging GPU acceleration through NVIDIA cuVS. These two achievements mark a significant advancement for retrieval augmented generation (RAG) search solutions. Continued development is underway to support larger-scale deployments beyond 4.8 billion vectors. Index build time on a massive-scale vector database is a crucial pain point for the industry. In collaboration with NVIDIA, Kioxia demonstrated up to 20x improvement in KIOXIA AiSAQ index build time for high-dimensional vectors of 1024 dimensions, and up to 7.8x improvement in end-to-end build times. This 20x improvement represents a reduction from 28.4 days using CPU to 1.4 days using four NVIDIA Hopper GPUs to build the index, and a reduction from 31 days to 4 days in end-to-end testing.[1] AI applications may now rely on larger volumes of vectorized information reaching tens of billions of vectors and beyond stored on SSDs, while DRAM alone becomes impractical even at a billion scale. Kioxia enables a highly scalable storage architecture with KIOXIA AiSAQ technology by achieving billion-scale search, exceeding RAG application latency requirements using a single query server in a Milvus vectorDB environment powered by GPU acceleration on index builds that make large scale deployments practical. "Vector databases provide a backbone for applications that need to understand intent, context, and similarity across massive, unstructured datasets in real time," said Jason Hardy, Vice President, Storage Technologies, NVIDIA. "By leveraging GPU-accelerated indexing with the NVIDIA cuVS library, Kioxia supports high-dimensional vector databases that can scale and build indexes with unprecedented efficiency." First announced last year, KIOXIA AiSAQ open-source software technology addresses RAG scalability challenges by enabling vector search directly from SSDs, with reduced DRAM usage. KIOXIA AiSAQ technology provides high scalability, making it well-suited for both multi-tenant environments and large-scale monolithic index deployments. The technology leverages an innovative Global Index algorithm that combines hybrid clustering and graph search to deliver efficient vector search at extreme scale. With flexible tuning options to balance performance and high-volume vector scalability, KIOXIA AiSAQ software makes large-scale deployments more accessible and easier to expand. "Scaling vector databases into the billions requires rethinking both memory and compute," said Masashi Yokotsuka, Managing Executive Officer, Vice President, SSD Division, Kioxia Corporation. "By combining KIOXIA AiSAQ SSD-based vector search with NVIDIA GPU acceleration for index construction, we provide practical index build at high scale deployments. As industry innovators, we will continue to push the boundaries of AI using flash memory." Kioxia remains committed to advancing storage-driven AI solutions that support intelligent data processing at scale and continues to evolve KIOXIA AiSAQ toward trillion-vector deployments. | Notes: | | 1. A total of 19.66 TB of vector data was processed for this benchmark. Performance or benchmark results may vary depending on the host device, read and write conditions, data sizes and other factors. | | KIOXIA AiSAQ is a trademark of KIOXIA. | | Company names, product names, and service names may be trademarks of third-party companies. | About Kioxia Kioxia is a world leader in memory solutions, dedicated to the development, production and sale of flash memory and solid-state drives (SSDs). In April 2017, its predecessor Toshiba Memory was spun off from Toshiba Corporation, the company that invented NAND flash memory in 1987. Kioxia is committed to uplifting the world with "memory" by offering products, services and systems that create choice for customers and memory-based value for society. Kioxia's innovative 3D flash memory technology, BiCS FLASH(TM), is shaping the future of storage in high-density applications, including advanced smartphones, PCs, automotive systems, data centers and generative AI systems. Information in this document, including product prices and specifications, content of services and contact information, is correct on the date of the announcement but is subject to change without prior notice. Contact details: Media Inquiries: Kioxia Corporation Promotion Management Division Satoshi Shindo Tel: +81-3-6478-2404

Nvidia GTC 2026: KIOXIA accelerates AI innovation with high-performance memory and SSD solutions. Showcasing AiSAQ Technology featuring CM9 Series Enterprise NVMe SSDs, super high IOs SSD emulator for AI-Ready storage, QLC storage for fast retrieval of large datasets featuring 245.76TB LC9 Series Enterprise NVMe SSDs. At Nvidia GTC 2026, Kioxia America, Inc. will showcase its latest high-performance memory and SSD solutions designed to meet the growing demands of AI workloads.As the premier global AI conference, GTC brings together developers, researchers, and business leaders to explore advances in accelerated computing and AI. At this year's event, the company will highlight the critical role of flash storage in enabling scalable, efficient AI infrastructures. By incorporating high-performance flash solutions, organizations can better optimize their AI investments, improve data access speeds, and accelerate time to insight. "GTC is the perfect venue to showcase how high-performance storage enables the next generation of AI applications," said Neville Ichhaporia, SVP and GM, SSD business unit, Kioxia America. "From accelerating AI-driven insights to improving infrastructure efficiency, our flash-based solutions help organizations translate data into actionable results with speed and reliability." Make AI Breakthroughs with Kioxia At booth #3522, located across from the main Expo Hall at the San Jose McEnery Convention Center, CA, Kioxia will demonstrate how its latest storage innovations support AI pipelines - from data ingestion and training to inference and analytics. LC9 Series Enterprise NVMe SSDs Product and technology demonstrations will take place March 16-19, and include: * Optimizing Memory Footprint While Improving Scalability for Vector Databases with Kioxia AiSAQ Technology - Featuring CM9 Series Enterprise NVMe SSDs * Super High IOs SSD Emulator for AI-Ready Storage - SSD emulator capable of delivering over 100 million IOs, designed to support demanding environments such as SCADA * High-Capacity QLC Storage for Fast Retrieval of Large Datasets - Featuring 245.76TB LC9 Series Enterprise NVMe SSDs * Kioxia Enterprise and Data Center SSD Portfolio - A comprehensive look at NVMe SSD solutions engineered for servers and storage systems, offering the performance, scalability, and reliability required for modern data center workloads Additionally, Rory Bolt, senior fellow and principal architect, Kioxia America, will present a theater session titled, 'Redefining Vector DB Management: Uniting Massive Scalability with GPU-Accelerated Indexing.' The session will take place in the Session Theater near the Kioxia booth on Wednesday, March 18 at 4:00 PM PT. Read also: Available in capacities of 512GB and 1TB, and package has been newly designed with a small 7.5 x 13 mm size Up to 1TB, with 4-bit-per-cell, QLC technology, QLC UFS boosts sequential write speeds by 25%, random read speeds by 90%, and random write speeds by 95% compared to previous-gen UHS-II SD memory card lines feature V90 and V60 video speed class ratings Continue to leverage AI-enabled smart manufacturing and economies of scale to ensure stable production of advanced 3D flash memory With PCIe 4.0, NVMe 2.0d specification compliant, 2230 and 2280 and 2242 form factor, and support for self-encrypting drive (SED) with TCG Opal version 2.01 Including Scalable BiCS FLASH 3D flash memory technology solutions, flash memory solutions for automotive, consumer and industrial applications, and broad lineup of SSDs Linus Tech Tips video provides behind the scenes look at Kioxia's Yokkaichi Fab production of BiCS FLASH 3D flash memory and SSDs SmartRAID 4300 accelerator supports up to 32 NVMe SSDs

Kioxia is working on a new high-speed SSD that extends a GPU's VRAM. Kioxia's Super High IOPS SSDs aim to deliver the high performance and low latency required for Nvidia's Storage-Next architecture. Mar 17, 2026, 12:44 GMT Kioxia has announced the development of an SSD that enable GPUs to directly access high-speed flash memory as an extension of their built-in High Bandwidth Memory (HBM). Kioxia says its Super High IOPS SSD is purpose-built to meet the growing demands of AI and high-performance computing, delivering efficient and low-latency data access. Made in collaboration with Nvidia's Storage-Next initiative, Kioxia's Super High IOPS SSD is described as a new type of SSD built for AI systems of today and the future, offering GPUs access to much more memory capacity than HBM chips, and without the high costs associated with DRAM products. This potentially makes it an ideal product to scale up data-intensive and AI systems while DRAM demand and prices are at an all-time high. Nvidia's Storage-Next initiative aims to address the anticipated shift from compute-focused to data-intensive workloads by expanding the GPU's access to memory spaces far beyond what's possible with GDDR7 and especially HBM. This should allow the GPU to work with larger data sets, which would improve its utilisation and latency by keeping data closer to the compute resources. The initiative calls on SSD vendors to design drives optimised for GPU-initiated workloads, effectively acting as an expansion of the GPU's packaged HBM memory. For its GP Series SSDs, Kioxia is using its high-performance XL-Flash memory, promising high IOPS, finer-grained 512-byte data access, and lower power consumption per operation, compared to conventional TLC NAND SSDs. Earlier reports indicate that the company is targeting 3x faster speeds than any available SSD with this NAND, and IOPS counts of 10 million. Large language models' Key-Value cache and GPU cache lines work with very small pieces of data, so SSDs optimised for many tiny random IOs, instead of large sequential transfers, can serve those caches much more efficiently. "Kioxia fully supports the Nvidia Storage-Next initiative and will deliver purpose-built SSDs to effectively address the need for GPU-accessible memory," said Makoto Hamada, senior director of the SSD Division, Kioxia Corporation. "This collaboration is instrumental in shaping the future of AI storage architecture." To accompany the new GP Series, the brand is also putting forward its CM9 Series PCIe 5.0 E3.S SSDs, which will support Nvidia's Context Memory Storage (CMX), and are designed to extend memory beyond the GPU. These will be available with 25.6TB of TLC capacity and 3 DWPD (drive writes per day) endurance. Kioxia has confirmed that evaluation samples of the GP Series will begin shipping in Q3 2026, with consumer availability by the end of 2026.