MinIO Plugs Apache Iceberg Tables Directly Into AIStor

Why It Matters

By consolidating data silos into a single high‑performance layer, MinIO enables faster AI model training and analytics while cutting cloud‑provider fees, reshaping enterprise data‑infrastructure economics.

Key Takeaways

• AIStor Tables GA with native Apache Iceberg V3 support
• Unifies structured and unstructured data via Iceberg tables
• Enables cost savings up to 40% versus AWS S3 Tables
• Deployable on-prem, private, sovereign, hybrid environments
• Provides Iceberg REST catalog views and multi-table transactions

Pulse Analysis

The rise of generative AI has amplified the need for a unified data layer that can serve both structured and unstructured assets at scale. MinIO’s AIStor Tables answers this demand by embedding Apache Iceberg’s open‑source table format directly into its object storage engine, eliminating the friction of moving data between databases and object stores. This native integration delivers SQL‑based access, versioned metadata, and atomic multi‑table transactions, allowing data scientists and analysts to query massive datasets without duplicating pipelines.

Technically, AIStor Tables distinguishes itself by implementing the full Iceberg V3 Catalog REST API, including support for Iceberg Views—virtual tables defined by SQL that capture query logic as metadata. The feature also leverages deletion vectors, row lineage, and geographic data types, positioning MinIO as a comprehensive analytics platform capable of exabyte‑scale performance. Because the tables run directly on the object store, latency is reduced and throughput is maximized, which is critical for real‑time model inference and large‑batch training workloads.

From a business perspective, the GA release offers enterprises a cost‑effective alternative to hyperscaler‑managed table services. With no extra per‑object fees and up to 40% storage savings, organizations can keep data on‑prem or in sovereign clouds while still accessing the same high‑performance analytics capabilities. This flexibility not only lowers total cost of ownership but also mitigates vendor lock‑in, making MinIO a compelling choice for companies seeking to accelerate AI initiatives without inflating cloud spend.

MinIO plugs Apache Iceberg tables directly into AIStor

AB Periasamy, co‑founder and CEO of MinIO

AIStor is MinIO’s object storage software and the Tables function refers to the open‑source Apache Iceberg software for large analytic tables accessible via SQL queries. Envisage these as a software layer above cloud object stores such as AWS S3, Azure Blob, GCP, plus formats like Parquet, ORC, and Avro, and now MinIO’s AIStor as well.

AB Periasamy, co‑founder and CEO of MinIO, said:

“Analytics and AI infrastructures are no longer defined by compute alone. The data layer now determines how much enterprise AI value can actually be realized.

When structured and unstructured data are unified, AI systems can learn more, reason better, and deliver greater impact. Only an object‑native architecture like MinIO AIStor can make that data fast, fluid, and ready for AI at scale. With AIStor Tables, we bring enterprise data together in a high‑performance data store that feeds analytics and AI systems directly.”

MinIO says the data‑for‑AI situation is simple but stark; AI systems have to work across fragmented and duplicated data pipelines because structured data lives in databases while unstructured data is held in object stores, to which we add files. But MinIO’s duplicated data pipelines point is still valid.

The AIStor Tables feature is a way to unify structured data and objects in a single data store using Apache Iceberg tables. That means there can be a single data pipeline hooking up analytic apps, AI agents, and models to an Iceberg‑using AIStor. MinIO says these data‑accessing functions can now “analyze structured and unstructured data together, operating on complete, up‑to‑date enterprise data at massive scale and performance.” It mentions exabyte‑scale.

MinIO claims it’s the first in the industry to build the full Apache Iceberg V3 Catalog REST API directly into the data store, providing Iceberg REST catalog views. These views are virtual tables defined by SQL queries that store metadata about the query definition. It says Apache Iceberg tables become first‑class citizens within AIStor, inclusive of views and multi‑table transactions.

MinIO’s AIStor Tables can be deployed across on‑premises, private, sovereign, and hybrid environments, reducing storage costs versus hyperscaler‑controlled table services, it claims, by up to 40 percent. That’s because the Tables feature is included in AIStor with no extra cost, which differs, it says, from AWS S3 Tables. AWS says that, with S3 Tables, you pay for storage, requests, and an object monitoring fee per object stored in table buckets. S3 Tables pricing can be seen here.

MinIO AIStor Tables is available today as part of MinIO AIStor. Enterprises can download and deploy directly from min.io.

Read about Iceberg Views in AIStor Tables in a blog, more about native Iceberg V3 here, and Iceberg deletion vectors, row lineage, geometry and geography types and more here.

---

Bootnote

The AIStor Tables feature entered limited tech preview status in September last year.