How Oracle’s New 23ai Database Brings AI-Powered Vector Search to Enterprises

Oracle is adding new artificial‑intelligence capabilities to its flagship database, renaming the product Oracle Database 23ai.

The company is delivering a long‑term‑support version called Database 23c, now marketed to enterprise users as Oracle Database 23ai.

The name change reflects the inclusion of AI‑related features that aim to simplify the development of AI‑driven applications and other major workloads.

Database 23c was first showcased at Oracle’s 2022 annual event, released to developers in early 2023, and then to enterprises, marking a strategic shift for the company.

Intense competition from other database vendors has pushed Oracle to refocus its strategy on developers, who can drive growth.

At the CloudWorld conference last September, Oracle announced plans to add vector‑search capabilities to Database 23c.

These AI vector‑search features introduce a new vector data type, vector indexes, and SQL operators, allowing the database to store the semantic content of documents, images, and other unstructured data as vectors and perform fast similarity queries.

Oracle also integrates Retrieval‑Augmented Generation (RAG) into 23ai, combining large language models with private business data to answer natural‑language questions.

Additional enhancements migrated from 23c to 23ai include JSON relational duality that unifies relational and document models, native graph queries on OLTP data, and JavaScript stored procedures, enabling developers to build applications using either relational or JSON paradigms.

Database 23ai will be offered as a cloud service and on‑premises through Oracle Exadata, Exadata Cloud@Customer, Oracle Base, and Oracle Database on Azure.

Pricing for 23ai has not been disclosed, but the developer version of Database 23c remains free to lower adoption barriers.

Oracle cites competitive pressure from vendors such as MongoDB, AWS, Google Cloud, Microsoft, Zilliz, DataStax, Pinecone, Couchbase, Snowflake, and SingleStore, all of which are adding AI‑oriented features like vector search.

Vector databases and vector search transform unstructured information into embeddings, making large‑scale storage, search, and comparison faster, more scalable, and more efficient.

Larry Ellison, Oracle co‑founder and CTO, emphasized that the new vector database lets customers augment foundation models with proprietary data, keeping that data private while enhancing AI performance for domains such as healthcare, finance, and personalized applications.