Introduction
At Oracle AI World 2025, one of the most exciting announcements came from the MySQL and HeatWave engineering team — the introduction of MySQL AI. This new innovation marks a turning point for data management, blending the power of Generative AI (GenAI), AutoML, and vector search directly inside the world’s most popular open-source database.
For decades, MySQL has been a trusted backbone for developers and enterprises alike. With over a billion installations worldwide and 190 million+ downloads annually, its impact is undeniable. Now, by adding native AI and ML capabilities, Oracle is transforming MySQL into a unified platform that bridges data storage, analytics, and intelligence — all within a single ecosystem.
The Evolution of MySQL — From Database to Intelligent Platform
For 30 years, MySQL has powered web applications, transactional systems, and enterprise workloads. Its proven reliability in LAMP (Linux, Apache, MySQL, PHP/Python) stacks and deep integration with cloud providers like Oracle Cloud Infrastructure (OCI), AWS, and Azure make it an industry favorite.
But traditional databases are evolving. Modern organizations deal with both structured data (invoices, medical records, orders) and unstructured data (PDFs, images, social content, and call logs). The modern data platform must handle both seamlessly — while providing natural language interaction, personalized insights, and real-time intelligence.
That’s exactly where MySQL AI steps in.
What Is MySQL AI?
MySQL AI is Oracle’s new initiative that embeds artificial intelligence directly into the MySQL ecosystem. It extends the capabilities of MySQL HeatWave, Oracle’s in-memory query accelerator, to include GenAI, AutoML, and vector processing — all while maintaining MySQL’s simplicity and efficiency.
Core Components of MySQL AI
- GenAI (Generative AI) — Enables natural language interfaces, content generation, summarization, and multilingual conversations.
- In-Database LLMs — Integrates large language models directly with MySQL data for context-aware question-answering.
- AutoML — Simplifies model creation, training, and deployment without requiring data science expertise.
- Similarity Search — Supports vector embeddings to compare and retrieve similar documents, images, or entries.
- Documents in File System — Enables querying of local documents without data migration.
- MySQL Studio — A centralized interface for managing AI models, data pipelines, and vector stores.
Together, these features redefine what “built-in intelligence” means for a database.
Why MySQL AI Matters
Oracle has designed MySQL AI to deliver AI/ML on-premises, leveraging existing hardware and ensuring high security for sensitive workloads. This approach eliminates the need for costly third-party tools or external data transfers.
Key benefits include:
- On-premise AI/ML — Leverage AI locally while retaining full control over data privacy.
- Document querying — Run intelligent searches directly over PDFs, TXT files, or CSV data.
- High security — Maintain encryption and access control under enterprise standards.
- Simplified deployment — No re-architecting or application changes required.
Seamless Integration with MySQL HeatWave
The power of MySQL AI builds on MySQL HeatWave, Oracle’s hybrid OLTP-OLAP engine that already supports analytics, machine learning, and now GenAI — all within the same MySQL service.
Migration is effortless:
- No changes to existing applications
- Lower cost and better performance
- Integrated analytics support
- Queries from object stores like OCI Object Storage, AWS S3, and Azure Blob
- Access to OCI GenAI Services (including Grok, Gemini, and Meta LLMs)
This architecture allows users to perform OLTP, analytics, AutoML, and GenAI on the same data — reducing latency and improving performance consistency.
Building GenAI Applications with MySQL AI
Developing AI-driven applications is now more accessible. MySQL AI provides a multi-step process for building GenAI apps using vector stores:
- Discover user documents
- Parse and extract metadata
- Split data into manageable segments
- Generate embeddings
- Insert vectors and metadata into a vector store
- Ensure ML model consistency for accurate querying
Once the knowledge base is built, developers can query it using vector search combined with LLMs to deliver contextual, human-like answers. This workflow simplifies Retrieval-Augmented Generation (RAG) within MySQL.
Performance: MySQL HeatWave Outpaces the Competition
According to Oracle’s benchmarks, MySQL HeatWave demonstrates 25× to 30× faster performance when creating vector stores compared to AWS Bedrock, and costs only one-fourth as much.
For similarity searches, HeatWave significantly outperforms BigQuery, Databricks, and Snowflake, offering superior price-performance ratios — a major advantage for data-intensive AI workloads.
Smarter Querying: Combining Semantic and Analytic Searches
MySQL AI introduces semantic search capabilities, allowing users to query data using natural language while simultaneously running analytic queries.
For example, in a Course Finder application:
- Semantic search retrieves course descriptions from a vector store.
- Analytic queries filter data by schedule, time, or instructor.
- Results are combined to deliver precise, context-aware recommendations.
This hybrid approach bridges structured SQL querying with unstructured, meaning-based retrieval — a game-changer for enterprise search.
Machine Learning and AutoML in Action
MySQL AI AutoML brings self-service ML directly to MySQL users through three key phases:
- Model — Create models automatically using AutoML.
- Deploy — Integrate ML models into applications via REST APIs or SQL.
- Gain Insight — Uncover hidden patterns, anomalies, and predictions.
This end-to-end cycle empowers teams to train, explain, and infer insights directly from their operational data.
Common Use Cases
- Customers: Segmentation, churn prediction, fraud detection.
- Products: Upsell/cross-sell, demand forecasting.
- Equipment: Predictive maintenance, root cause analysis.
- Financials: Credit risk assessment, payment prediction.
By uniting AutoML and GenAI, MySQL AI allows for synergistic value creation — leveraging AutoML for data filtering and GenAI for explanation and reasoning.
Vector Operations — The Heart of MySQL AI
MySQL AI introduces native vector data types and distance functions such as:
VECTOR_DISTANCE(vector, vector, distance_metric)STRING_TO_VECTOR(string)VECTOR_TO_STRING(vector)
Supported distance metrics include:
- Cosine
- Dot product
- Euclidean
- Manhattan
These operations are parallelized and in-memory, ensuring high performance. A vector index is also available, supporting scalable similarity search — the foundation for applications like recommendation engines and document retrieval.
Natural Language to SQL (NL2SQL)
One of the most impressive features of MySQL AI is NL2SQL — a natural language interface that translates user questions into SQL queries automatically.
For instance, a business executive can ask:
“Which products have the lowest cost per click?”
The system retrieves schema metadata, uses an LLM-powered augmented prompt, validates SQL syntax, and executes the query — all seamlessly.
This capability dramatically reduces dependency on SQL experts and democratizes data access for non-technical users.
Conclusion: A New Era for Intelligent Databases
With MySQL AI, Oracle has redefined the boundaries of what a modern database can achieve. By integrating GenAI, AutoML, and HeatWave analytics, MySQL now delivers end-to-end intelligence for every data-driven enterprise. Whether you’re managing transactional workloads, running analytics, or building AI-powered applications, MySQL AI empowers you to innovate faster, smarter, and more securely.
For readers who want to explore Oracle’s broader vision for intelligent data platforms, check out:
🔗 Oracle AI Database 26ai — The Future of AI-Driven Data
🔗 Inside Oracle AI World 2025: Oracle AI Database 26ai and the Rise of the AI DBA
These articles dive deeper into how Oracle’s AI ecosystem — including MySQL AI — is shaping the next generation of autonomous, intelligent databases.
