Oracle Database 23ai is the long-term-support release that follows 19c. It carries roughly 300 new features, but only a handful matter for licensing. The licensing question is not which features Oracle has added — it is which features pull in Database Options that the customer was not previously paying for. Oracle's standard upgrade path from 19c to 23ai is licence-neutral on paper: existing Enterprise Edition and Standard Edition 2 licences transfer to 23ai under the support contract without re-licensing. In practice, several of the headline 23ai features touch the Database Options framework, and the upgrade can quietly create option exposure that LMS will reconcile at the next audit. AI Vector Search is the most-discussed example, but it is far from the only one. This guide walks through every major 23ai feature, classifies it as licence-neutral, option-triggering, or contract-requiring, and explains the redline moves to apply before scheduling the 19c-to-23ai cutover.
Oracle Database 23ai is positioned as a free upgrade for existing 19c (or earlier supported releases) customers under their support contract. The base licence transfer is unconditional — Enterprise Edition stays EE, Standard Edition 2 stays SE2, Named User Plus and Processor metric counts carry across. There is no re-licensing fee. The standard upgrade path is technically clean and licensingly clean as long as the customer adopts only the features that ship inside the base edition.
That last clause is where the work lives. Oracle has packaged some of 23ai's most marketed features inside the base edition, and some inside the Database Options framework. The classification is not always obvious from the marketing pages. See the Database Licensing Guide for the underlying options model. The classification below walks each major 23ai feature through its licensing impact.
This list is the safe set. Adopting any of these features alone, without other architectural changes, carries no incremental licence cost on top of the existing 19c entitlements. The trap is not in the features themselves; it is in the secondary architectural decisions that often accompany their adoption.
Some 23ai features are themselves part of the Options framework, or they pull in Options when used at production scale. Adopting these in production without licensing the underlying Option creates audit exposure.
| 23ai feature | Triggers Option | Option list price |
|---|---|---|
| Operational Property Graph (production scale) | Spatial and Graph | $17,500/proc |
| Globally Distributed Database | Sharding (separate, on top of EE) | ~$23,000/proc estimated |
| True Cache (full features) | Active Data Guard | $11,500/proc |
| SQL Tuning Sets / Tuning Advisor on 23ai workload | Tuning Pack | $5,000/proc |
| AWR / ADDM on 23ai workload | Diagnostics Pack | $7,500/proc |
| Compressed segments on vector / JSON Duality tables | Advanced Compression | $11,500/proc |
| Partitioned vector or JSON indexes | Partitioning | $11,500/proc |
| In-memory column store on 23ai schema | Database In-Memory | $23,000/proc |
The pattern that catches teams: adopting AI Vector Search (free) plus Partitioning (paid) plus Diagnostics and Tuning (paid) creates $30K+ per processor of net-new option exposure that the upgrade plan did not flag. See RAG on Oracle Database licensing for the worked option footprint.
A small number of 23ai features sit outside the standard Options framework entirely and require new contract terms before they can be used. The most prominent is the AI Vector Search foundation-model integration with Oracle Generative AI Service — the Vector Search SQL is free, but the GenAI Service inference layer is metered and needs an ordering document. The same applies to Select AI when wired to Oracle Generative AI Service.
Globally Distributed Database (the rebranded sharding stack) is technically available in 23ai but requires a Sharding option ordering document. Operational Property Graph at large scale needs the Spatial and Graph option in writing. Adopting these without ordering documents in place is the cleanest path to a LMS-driven repricing at next audit.
For a worked example of a 19c-to-23ai migration with full option footprint analysis, the License Optimization service runs the assessment and produces the redline package before contract execution. The broader audit-defence pattern is covered in the Oracle audit guide.
Independent, buyer-side analysis. Fixed-fee, 10 business day turnaround. Former Oracle insiders, 25+ years, $1.8B in Oracle spend advised.
No - the database licence transfers under the support contract. What can change is the Database Options footprint: if a 23ai feature you adopt uses an option you previously did not license, that is net-new exposure even though the base licence transfer is free.
Yes - AI Vector Search is included in Enterprise Edition and Standard Edition 2 at no incremental licence cost. The risk is that AI Vector Search is usually deployed alongside Partitioning, Compression, Diagnostics and Tuning, which are options.
JSON Relational Duality is a 23ai feature that lets developers operate on relational tables as if they were JSON documents. It is included in EE and SE2 at no extra cost, with no option dependency.
Select AI is included in 23ai EE for the SQL syntax that pipes results to an LLM. The LLM itself is billed separately: through Oracle Generative AI Service if used through OCI GenAI, or through BYO key if used with a third-party LLM. The Select AI feature itself is free.
Option triggering through 23ai feature adoption. The biggest single risk we see is teams adopting JSON Relational Duality plus Partitioning to manage scale - and Partitioning is a separately licensed option ($11,500 per processor list).
Twice a month. Oracle pricing moves, audit-defence tactics, GenAI Service rate changes. Written by former Oracle insiders.
No spam. Unsubscribe any time. Independent — not affiliated with Oracle Corporation.