MinIO, the data and memory foundation for enterprise AI, today announced at SuperComputing 2026 (SC26) the general availability of AIStor Memory, alongside product and partnership milestones marking AIStor’s second anniversary.
As AI unlocks new value in enterprise data, organizations need to bring all their data together and capture more of the information their businesses generate. Since its November 2024 launch, AIStor has evolved to meet that need, expanding from objects to native Apache Iceberg tables and persistent agentic memory. This evolution enables enterprises to build a sovereign AI lakehouse where structured and unstructured data is secure, governed, and available for AI and analytics, while preserving the knowledge agents generate on infrastructure they control.
“Enterprise AI is moving into the workflows that run the business. Agents need access to enterprise data, persistent context, and the ability to build on prior work. In two years, AIStor has brought agent memory, tables, files and objects onto one foundation, with customers operating at exabyte scale,” said AB Periasamy, co-founder and CEO of MinIO. “With AIStor Memory now generally available, enterprises can turn what agents learn into lasting organizational knowledge. That knowledge belongs to the enterprise, on the infrastructure it controls.”
Building a Sovereign AI Lakehouse
AIStor’s expansion from objects to native Apache Iceberg tables enables organizations to build a lakehouse on infrastructure they control, rather than creating a separate data platform for every application.
Within the lakehouse, Iceberg tables can index logs, telemetry, financial transactions, machine-generated events, and vector embeddings, alongside information from documents, audio, and video. A common catalog makes that data discoverable to analytics tools and authorized AI agents, connecting enterprise data modernization directly to AI adoption.
The same lakehouse supports observability by bringing operational logs and events together for ongoing analysis. AIStor Memory adds persistent context, preserving what agents learn as governed organizational knowledge they can build on across sessions. Enterprises retain control of both the data that powers AI and the knowledge it generates.
Expanding the Data and Memory Foundation
Since launch, AIStor has added S3 Express API support, S3 over RDMA, and ExaPOD, an AIStor-powered reference architecture developed with Intel, Solidigm, and Supermicro for exabyte-scale deployments. In its second year, it extended that foundation with native tables, direct enterprise data sharing, and durable organizational memory.
Key innovation milestones from the past year include:
- AIStor Tables: Generally available in February 2026, AIStor Tables unifies native Apache Iceberg tables and objects in the same data store. It gives enterprises a common foundation for organizing data, making it discoverable through a catalog and supporting analytics and agentic AI without maintaining separate platforms for tables and objects.
- AIStor Open Sharing for Databricks: Introduced in March 2026, AIStor’s native sharing capabilities connect on-premises enterprise data directly to Databricks. Customers can apply analytics and AI to data where it resides without first creating and maintaining duplicate cloud datasets or separate replication pipelines.
- AIStor Memory: MinIO today announced the general availability of AIStor Memory, following its July 2026 technology preview. AIStor Memory unifies long-term memory, persistent workspaces, and secrets on enterprise-controlled infrastructure, giving enterprises a production-ready foundation for agents to retain context across sessions and build on prior work. It preserves what agents learn as governed organizational knowledge that authorized agents can reuse, helping improve response quality while reducing latency and token costs.
Together, these capabilities address two requirements for production enterprise AI: making business data accessible to agents and preserving the knowledge those agents generate.
“Enterprises need AI to work with their data without creating another layer of cost and complexity,” said Garima Kapoor, co-founder and co-CEO of MinIO. “Our customers bring data, analytics and agent memory onto one foundation instead of rebuilding their infrastructure for every new workload. That gives them more room to invest in what AI can do for their business.”
Global Partnerships Extend MinIO’s Role in Enterprise AI
MinIO’s collaborations across accelerated computing, data platforms, sovereign AI and observability extend AIStor’s role in the environments where enterprise data resides.
- NVIDIA: MinIO joined the NVIDIA STX ecosystem in March 2026 and announced additional work in May to support the NVIDIA Vera BlueField-4 STX storage architecture. The collaboration brings AIStor’s persistent data foundation together with NVIDIA accelerated computing to support secure, high-performance enterprise AI factories.
- Databricks: MinIO was named the 2026 Databricks ISV Storage Partner of the Year, recognizing the companies’ work helping enterprises apply analytics and AI to governed data where it resides. Their collaboration enables customers to extend Databricks workloads to on-premises datasets without building separate replication infrastructure and provides a foundation for additional capabilities the companies are developing together.
What Our Partners Are Saying
“Our customers want to use Databricks across all their data, including data that must remain on-premises. With OpenSharing built into AIStor, teams can query on-premises Iceberg and Delta tables and govern them in Unity Catalog, without replication pipelines to build or maintain,” said Stephen Orban, SVP, product ecosystem & partnerships at Databricks. “That progress earned MinIO our 2026 ISV Storage Partner of the Year award. We value our partnership with MinIO and its commitment to OpenSharing, which gives our joint customers an open, governed way to put on-premises data to work for analytics and AI.”
“Infrastructure efficiency becomes increasingly important as enterprise AI scales. Our testing with MinIO AIStor and Micron 6600 ION SSDs demonstrates how an optimized storage architecture can deliver greater performance while reducing demands on power and compute resources,” said Larry Hart, senior director of solutions marketing for Micron’s Core Data Center Business Unit. “That efficiency helps free system resources to support the growing demands of AI.”
Availability
AIStor, including AIStor Tables and AIStor Memory, is now generally available.
Resources
- MinIO AIStor
- AIStor Tables: Technical Deep Dive
- AIStor Open Sharing
- AIStor Memory
- Case Study: AMD data intelligence on MinIO AIStor
- Case Study: Nomura replaced HDFS with MinIO AIStor
About MinIO
MinIO is the data and memory foundation for enterprise AI. Built for the speed, scale, and economics of modern AI workloads, AIStor unifies objects, tables, and agentic memory, while MemKV provides a dedicated foundation for inference context. Spanning core, edge, and cloud, MinIO powers AI factories, intelligent applications, and autonomous agents with the performance, governance, and control enterprises need to build production AI at scale. Trusted by 77% of the Fortune 100, MinIO is backed by AME Cloud Ventures, Dell Technologies Capital, General Catalyst, Index Ventures, Intel Capital, SoftBank Vision Fund 2, and other leading investors.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261118649010/en/
“Enterprises need AI to work with their data without creating another layer of cost and complexity,” said Garima Kapoor, co-founder and co-CEO of MinIO.
Contacts
Media Contact
press@min.io
