MariaDB Unveils Single Database for Enterprise AI
MariaDB Enterprise Platform 2027 is now generally available, and its AI Retrieval service is in public preview on MariaDB Cloud.
Updated on Nov. 4, 2026 in Artificial Intelligence
MariaDB has launched MariaDB Enterprise Platform 2027 as a single database for enterprise AI agents, now generally available. Its AI Retrieval service, which turns natural-language questions into read-only SQL, entered public preview on MariaDB Cloud.
Why it matters
MariaDB states that every integration between separate systems adds latency, synchronization lag, and a point of failure, which the platform aims to reduce by consolidating agent memory, caching, analytics and vector search in one system. The platform's FIPS 140-3 certification and Bring Your Own Cloud option address enterprise requirements over where data is kept.
Native vector search delivers up to 30 percent more queries per second than the last major release, and peak throughput is 56 percent higher in third-party testing. The source does not disclose the test workloads or the testing party's methodology.
The players
MariaDB
MariaDB is an open-source database company whose products are widely used in enterprise data systems.
The details
The platform bundles Agent Memory, a durable in-memory store shared across agent sessions, and Semantic Cache, which returns stored answers to questions phrased differently. MariaDB Exa runs real-time analytics on live operational data, while Raft clustering and conflict detection and resolution support multi-region deployments.
Timeline
• November 4, 2026: MariaDB Enterprise Platform 2027 announced and made generally available.
The Tech Race
The launch positions MariaDB against rival database vendors by claiming gains over its own last major release in vector search and peak throughput. It reflects a broader push to fold AI agent memory, retrieval and analytics into the operational database rather than separate systems.
Teams building AI applications could reduce the number of separate systems they must integrate for memory, caching, retrieval and analytics. Organizations with data-residency rules may use the Bring Your Own Cloud option to keep data where they choose.
The takeaway
Fewer integrations between separate systems can mean fewer points of failure and less synchronization lag for AI-driven applications. Developers evaluating such platforms should test performance claims against their own workloads.
Further reading
Explore more on Artificial Intelligence developments shaping enterprise software.
