What Oracle 23ai Actually Is — and Why Licensing Is More Complicated Than the Marketing Suggests
Oracle Database 23ai (the renamed successor to 23c) is Oracle's first database release to position AI Vector Search as a native, first-class feature. Oracle's marketing positions AI Vector Search as included in all 23ai editions — including Standard Edition 2 and the free Oracle Database Free edition. This is largely accurate for basic vector storage and similarity search. What Oracle's marketing does not prominently explain is that the AI features most enterprises actually want — the integrations with large language models, the Select AI natural language query interface, and the AI Profile configuration framework — carry separate licensing requirements that can substantially increase your Oracle bill.
This guide breaks down exactly which 23ai AI capabilities are genuinely included versus which trigger additional licence costs. For the broader Oracle licence management picture, our Oracle Knowledge Hub is the starting point, and our Oracle advisory team works specifically on 23ai licensing reviews for organisations mid-migration from 19c.
What Is Included in Standard 23ai Licences
AI Vector Search — the ability to store, index, and query vector embeddings directly in Oracle Database — is included in all paid Oracle Database editions (Enterprise Edition, Standard Edition 2) and in Oracle Database Free. The VECTOR data type, VECTOR_EMBEDDING and VECTOR_DISTANCE functions, and HNSW and IVF vector indexes are all base-product features requiring no additional option licences.
This is genuinely significant: organisations that need to run Retrieval Augmented Generation (RAG) pipelines by storing embeddings in Oracle and performing vector similarity search no longer need a separate specialised vector database. The operational and licensing simplification for organisations already standardised on Oracle is real. For organisations evaluating Oracle's cloud direction alongside 23ai, our Oracle Alloy licensing guide covers how 23ai runs in Alloy-based cloud environments.
Assess Your Oracle 23ai Licence Position
Use our Oracle licence assessment tools to map your current 19c/21c options against 23ai requirements before migrating.
Start Free Oracle Assessment →What Triggers Additional Licence Costs
Select AI — Oracle's natural language to SQL interface that lets users query the database using plain English — requires Oracle Autonomous Database or specific OCI services. It is not available as a standalone on-premises option and therefore requires a move to OCI or an OCI-based Autonomous Database deployment. For organisations evaluating Select AI, this is a cloud migration decision as much as a licensing decision.
AI Profile and DBMS_CLOUD_AI integration — the PL/SQL package that connects Oracle Database 23ai to third-party LLM providers (OpenAI, Azure OpenAI, Google Gemini, Cohere) for in-database AI inference — requires connectivity to OCI or to OCI Object Storage. On-premises 23ai deployments require specific network and credential configurations that, while technically possible, are not officially supported in all deployment topologies. Oracle's sales team will typically position OCI-based deployments as the supported path.
| Feature | Included In | Additional Licence Required? |
|---|---|---|
| VECTOR data type | All 23ai editions | No |
| AI Vector Search (HNSW/IVF) | All 23ai editions | No |
| VECTOR_EMBEDDING function | All 23ai editions | No (model must be local) |
| Select AI (NL to SQL) | Autonomous Database (OCI) | Requires OCI / Autonomous DB |
| DBMS_CLOUD_AI (LLM integration) | OCI-connected deployments | Requires OCI or Exadata Cloud |
| AI Profile configuration | OCI-connected 23ai | OCI dependency; token costs apply |
| In-database ONNX model execution | Enterprise Edition | No (EE only — not SE2) |
Planning a Migration from Oracle 19c or 21c to 23ai?
Our Oracle advisory team conducts 23ai licensing reviews as part of migration planning — identifying which AI features require OCI, which trigger option licences, and how AI feature usage affects ULA certification counts.
Talk to an Oracle Specialist →How 23ai AI Features Affect ULA Certification Counts
For organisations operating under an Oracle Unlimited Licence Agreement (ULA), the AI Vector Search and related 23ai features deployed under the ULA must be counted and certified at ULA exit. Oracle's position is that any deployment of a licenced product — including deployments specifically to support AI features — counts toward the certification metrics (typically Named User Plus or Processor). Organisations that deploy 23ai broadly across their estate to enable AI Vector Search for multiple applications may find that this substantially increases their certified processor count at ULA exit.
This is a particular risk for organisations mid-ULA who were planning a relatively contained certification. If your ULA is approaching expiry and you are deploying 23ai for AI workloads, our Oracle ULA guide explains the certification mechanics in detail. Our advisory team has worked through several ULA certifications where 23ai AI deployments created unexpected count inflation — and the time to address this is before certification, not after. Book a confidential call to discuss your specific position. For related Oracle cloud product topics, see our Oracle NetSuite pricing guide and Oracle Alloy guide.