MinIO AIStor Brings Object Data Stores for the NVIDIA STX Reference Architecture

AIStor provides a unified data foundation supporting the NVIDIA STX reference architecture, accelerating training, enterprise RAG, and real-time agentic inference throughout the AI lifecycle

‍

REDWOOD CITY, Calif., March 16, 2026 — MinIO, the data foundation for enterprise analytics and AI, today announced that MinIO AIStor will support object data stores for the NVIDIA STX reference architecture. Designed with the NVIDIA STX rack-scale reference architecture, AIStor delivers a unified, high-performance datastore that powers the full AI lifecycle—from large-scale model training to enterprise RAG and real-time agentic inference. As part of this collaboration, MinIO is joining the NVIDIA STX ecosystem as a partner for AI data platforms.

NVIDIA STX defines how enterprise AI factories store, move, and access data at modern AI speeds. Powered by NVIDIA Vera Rubin, NVIDIA BlueField-4 processor, and NVIDIA Spectrum-X Ethernet networking, STX provides a modular, rack-scale storage architecture for production AI infrastructure. Running on NVIDIA STX reference architecture, AIStor serves as an object-native data foundation at the center of the AI factory.

As AI systems evolve from isolated inference to distributed, multi-agent workflows, infrastructure demands expand across training throughput, RAG indexing performance, and low-latency context access. AIStor addresses these requirements within a single, consistent platform that operates with the NVIDIA STX architecture.

MinIO: One Data Store for the Full AI Lifecycle

High-Performance Data Store for Training and Analytics

• Leveraging NVIDIA STX reference architecture, AIStor delivers optimal throughput and power density required to keep GPU clusters fully utilized and energy efficient. Its fully disaggregated design aligns with STX’s rack-scale model, separating storage from compute to enable independent scaling and fault isolation without compromising low-latency performance.
• Running locally on the BlueField-4 processor, AIStor is designed to saturate 800GbE, delivering the full network bandwidth of the platform directly to AI workloads without a host CPU in the path.
• AIStor unifies data and metadata within a distributed namespace that scales to multiple exabytes, eliminating the need for centralized metadata tiers and performance bottlenecks as deployments grow.

Enterprise AI Data: RAG and Multimodal Indexing 

• Built on STX architecture, AIStor provides a data foundation for RAG and multimodal indexing pipelines.
• Running natively on NVIDIA BlueField-4 processors as part of the STX design, AIStor accelerates embedding, indexing, and retrieval at wire speed while preserving S3 API compatibility, reducing latency and operational complexity compared to traditional storage tiers.

Context Memory for Agentic AI 

• As AI advances from single-turn prompts to multi-step reasoning systems, the bottleneck shifts from compute to sustained context memory. NVIDIA STX provides the foundation for the NVIDIA Context Memory Storage platform (CMX), and AIStor extends the memory hierarchy beyond GPU limits, together forming the rack-scale infrastructure required for these workloads.
• Modern models generate large volumes of key-value (KV) cache data during inference. Within STX, AIStor delivers scalable, high-performance object storage to retain and retrieve inference state with near-DRAM predictability across distributed agents.

“AI is being reshaped by data as much as by models,” said Garima Kapoor, Co-CEO and Co-Founder of MinIO. “The infrastructure that succeeds in this era won’t be retrofitted for AI; it will be architected for it. AIStor’s integration of the NVIDIA STX reference architecture reflects years of deliberate engineering around performance, scale, and architectural simplicity. We are delivering the unified, object-native data foundation that production AI factories require.”

MinIO AIStor: Architected for NVIDIA STX
AIStor’s alignment with NVIDIA STX reflects deliberate architectural decisions made well before rack-scale AI factories became mainstream. This includes standardizing on ARM and ensuring that MinIO’s minimal footprint integrates natively with the NVIDIA BlueField-4 processor that underpins the STX architecture.

AIStor ships as a single, static binary under 200 MB, with no external metadata databases or background services. Its lightweight design runs natively within BlueField-4, which underpins the STX storage tier, eliminating dedicated x86 storage nodes and reinforcing STX’s disaggregated architecture.

MinIO’s long-standing ARM64 and Scalable Vector Extension (SVE) optimizations further strengthen this fit. AIStor’s erasure coding using ARM SVE delivers up to 2× faster throughput compared to earlier implementations, while integrity verification scales linearly with core count. These optimizations position AIStor as a natural storage layer for the silicon foundation of NVIDIA STX.

STX-Enabled Capabilities
Within the NVIDIA STX architecture, AIStor unlocks advanced capabilities through NVIDIA DOCA and NVIDIA BlueField-4 acceleration:

• NVIDIA GPUDirect RDMA for S3-Compatible Storage — Sub-millisecond object access (tech preview)
• Hardware-Accelerated Erasure Coding — Wire speed SVE2 SIMD accelerated erasure coding and encryption tolerance offloaded to BlueField-4
• Zero-Copy Data Transfer — Direct movement from NVMe to GPU memory
• Hardware-Accelerated Encryption, Checksums, and Compression — Inline data protection and optimization

Together, these capabilities ensure storage operates as an integrated component of the STX  rather than an external bottleneck.

Availability

MinIO AIStor on NVIDIA BlueField-4 within the NVIDIA STX reference architecture is expected to reach general availability in the second half of 2026, with early access available for qualified enterprise customers. AIStor with NVIDIA GPUDirect RDMA for S3-compatible storage is available now as a tech preview. SVE2 and ARM64 optimizations are production-ready today on BlueField-3 and other ARM platforms.

Resources

• Blog: Benchmarking Vector Index Creation with MinIO AIStor, Milvus, and NVIDIA cuVS
• Blog: AIStor Runs Natively Inside NVIDIA BlueField-4: Object Data at Wire Speed 
• Whitepaper: MinIO AIStor: The Unified Data Foundation for the NVIDIA AI Factory

About MinIO
MinIO is the data foundation for enterprise analytics and AI. Built for exascale performance and limitless scale, MinIO AIStor delivers a secure, sovereign, and AI-ready data store that spans from the edge to the core to the cloud. With rampant adoption across the Fortune 100 and 500, MinIO is redefining how organizations and government agencies store, manage, and mobilize all of their data in the AI era. MinIO is backed by Jerry Yang's AME Cloud Ventures, Dell Technologies, General Catalyst, Index Ventures, Intel Capital, Softbank Vision Fund 2, and others.

Media Contact:
Jennifer Marquez
jennifer.marquez@archetype.co

‍