MinIO has announced MemKV, a context memory store designed for AI inference workloads that delivers microsecond context retrieval at petabyte scale. Built for NVIDIA STX architecture, MemKV addresses the "recompute tax" caused when AI systems lose context and must repeat completed work. The company claims MemKV substantially reduces time-to-first-token at production concurrency. In a typical enterprise deployment with 128 GPUs and 128K-token context length, the technology increased GPU utilisation from approximately 50% to over 90%, resulting in $2 million in annual compute savings. MemKV is the second product in MinIO's portfolio alongside AIStor, extending the company's data foundation into the memory tier where inference runs. The system provides persistent, shared context across GPU clusters, breaking the traditional tradeoff between high-speed memory and scalable storage.

Four Inc. has been named the public sector technology provider for MinIO, the data foundation for enterprise analytics and AI. The partnership will make MinIO's AIStor cloud-native object storage platform available to government agencies through Four Inc.'s NASA SEWP V and ITES-SW2 contract vehicles. MinIO's software-defined platform unifies structured and unstructured data into a single namespace, designed to handle AI training pipelines and data-intensive workloads whilst meeting federal security standards. The system is built for exascale performance and can operate across edge, core and cloud environments. The collaboration combines Four Inc.'s public sector expertise with MinIO's data infrastructure capabilities. MinIO is backed by investors including AME Cloud Ventures, Dell Technologies, General Catalyst and Softbank Vision Fund 2.

MinIO has announced that MinIO AIStor will support object data stores for the NVIDIA STX reference architecture. The collaboration positions AIStor as a unified, high-performance datastore designed to power the full AI lifecycle, from large-scale model training to enterprise RAG and real-time agentic inference. Built on NVIDIA STX's rack-scale architecture, which includes Vera Rubin, BlueField-4 processor and Spectrum-X Ethernet networking, AIStor operates as an object-native data foundation for AI factories. Running locally on the BlueField-4 processor, AIStor is designed to saturate 800GbE, delivering full network bandwidth directly to AI workloads. The platform features a distributed namespace scaling to multiple exabytes, eliminating centralised metadata bottlenecks. MinIO is joining the NVIDIA STX ecosystem as a partner for AI data platforms.