Oracle Autonomous AI Database

Use it when

• The estate is already Oracle; PL/SQL, existing ETL, and DBA-shaped skills carry over without a migration project.
• You want the operational surface gone: provisioning, tuning, patching, backup, and recovery are the service's job, with auto-indexing standing in for hand tuning.
• Deployment flexibility matters: shared serverless, dedicated Exadata, Cloud@Customer in your data center, and Oracle Database@ inside Azure, Google Cloud, and AWS.
• Two Always Free instances make indefinite evaluation genuinely free.

Think twice when

• You are not an Oracle shop; greenfield analytics teams usually land on Snowflake, BigQuery, or a lakehouse first, and Oracle SQL plus PL/SQL is its own world.
• Cost transparency matters; ECPU pricing is straightforward, but the surrounding platform decisions (OCI, support, licensing culture) are part of what you are buying.
• Open table formats should be the primary store; it reads Iceberg, Parquet, and Delta but stores proprietary.

How it runs

Fully managed on Exadata: shared-disk architecture where ECPUs scale independently of storage, per-second billing, auto-scaling to three times baseline, and scale-to-zero when idle on serverless. Predefined HIGH/MEDIUM/LOW services shape concurrency. GoldenGate and Kafka connectors feed it; Flashback provides 60-day time travel by default.