Oracle plies its unique architecture to step on the generative AI accelerator

Omdia view

Summary

At its annual CloudWorld user conference held in Las Vegas Nevada last month, Oracle placed its foot squarely on the generative AI (GenAI) accelerator with several announcements that leverage the company’s full stack Oracle Cloud Infrastructure (OCI) architecture and its unique ability to bring AI and data closer together. In this report, Omdia will dive into two key announcements to better understand how the company is evolving to tackle AI, multi-cloud, and customer success using and in support of GenAI technologies.

Tearing down the hyperscale walled garden

Given the scope of Oracle CloudWorld, which hosted more than 15,000 customers and partners, it is easy to lose sight of the big picture amid the typical onslaught of major introductions (more than twenty this time around) ranging from embedding more than 100 AI capabilities into Oracle Fusion Cloud business applications to running Oracle Database on OCI within AWS data centers.

The latter of which, dubbed Oracle Database@AWS, understandably garnered the most attention during the conference. The idea of enabling Oracle Autonomous Database on dedicated infrastructure and Oracle Exadata Database Service within AWS data centers is very compelling. Oracle Database users, for example, can now connect directly with AWS apps and tools such as Amazon SageMaker and Bedrock, all without incurring data ingress and egress charges.

This new service is scheduled to be available in preview later this year with limited availability across AWS regions and broader availability in 2025. But it is not alone. Oracle Database@AWS joins two similar efforts, Oracle Database@Azure and Oracle Database@Google Cloud. Considered together, these new offerings put Oracle in a unique and advantageous position. The company is not merely selling subscriptions running on these competing cloud platforms, hosting Oracle Database within a basic Kubernetes container running on Microsoft Azure Linux, for example.

Rather, Oracle is offering its full hardware/software stack (OCI) on AWS, a combination that enables unique functionality such as autonomous database management, seamless access to AWS services like Amazon S3, and integrated systems tooling across both with Oracle services represented natively within tools such as AWS CloudFormation and AWS Management Console.

In this way, Oracle looks a bit less like a partner and more like the infamous cuckoo bird, which lays its eggs in the nests of its rivals (warblers, finches, etc.) with the intent of displacing rival chicks to more effectively compete for food and shelter. In Oracle’s case, the company now offers unique capabilities such as hybrid/multi-cloud deployments and in-database AI inference for AWS, Google, and Microsoft cloud customers right where they live, all without compromise.

This benefits existing Oracle Database customers, seeking to better incorporate cloud services across Google, AWS, and Microsoft. It helps, of course, that Oracle has established interconnect deals with both Google and Microsoft, which does away with the cost of moving data in and out of the Oracle database and establishes single-pane management for services from Oracle and the host cloud platform.

But Oracle’s new trio of hyperscale database options also presents non-Oracle customers with some interesting opportunities. By tearing down the walls that typically define and separate each hyperscale platform, Oracle enables users to very easily pivot from established databases like AWS RedShift and Google BigQuery to instead adopt Oracle Database in order to take advantage of the company’s converged data architecture that supports in-database AI, all modern data types, workloads, and development styles, complemented with automated database management and hybrid/multi-cloud capabilities.

As but one example, perhaps it is advantageous to build a retrieval augmented generation (RAG) pipeline using Oracle Database 23ai's efficient vector search facilities, feeding those search results to a GenAI app built on Google Vertex AI. Customers now have the option to build this kind of multi-cloud solution but do so right on Google Cloud.

There can be no AI without data

As mentioned earlier, with major announcements like Oracle Database@AWS garnering the majority of attention, it is easy to overlook smaller but potentially more important shifts in Oracle’s AI, analytics, and data portfolio itself. For example, Oracle is currently making its own pivot to better support semi-structured and unstructured data with the announcement of the Oracle Intelligent Data Lake.

Billed as an evolution of Oracle OCI Data Lake, this new offering will subsume the company's data catalog, data integration (Oracle Data Flow), and data lake services, melding them into a single offering, which will be available in 2025. As you would expect from a data lake, it will offer capabilities to connect, extend, inventory, transform, and orchestrate data, all using a unified governance and security.

And in keeping with current market trends, Oracle’s implementation will focus on interoperability through the adoption of open table formats, most notably Apache Iceberg and Apache Parquet. Likewise, it will incorporate a flexible compute engine, supporting key open source distributed data processing frameworks, Apache Spark and Apache Flink; and it will come with its own modern implementation of Jupyter notebooks next to a no-code user experience.

Next to its competitors, Oracle’s new data lake looks like a following move, given that data lakes and data lakehouses have become quite common over the last five years. More recently, with the sudden rise of GenAI and its hunger for unstructured data, data lakes have become wholly indispensable—not just to store raw data but to also serve as a central repository of corporate knowledge, supporting both AI and analytics workloads.

This is the key to Oracle’s planned re-introduction of a data lake. Oracle is not playing catch up but rather constructing a comprehensive, portfolio-wide data layer, one that sits Oracle Intelligent Data Lake alongside Oracle Autonomous Data Warehouse, Oracle Analytics Cloud, Oracle HeatWave, as well as several AI services and third-party partner services.

What’s next?

In this way, Oracle can position OCI and all of its attendant database services as a complete data platform, something Oracle is already referring to as the Oracle Data Intelligence Platform. Not merely a unified platform in name alone with each product operating in isolation, Oracle's Data Intelligence Platform will actually establish a unified data layer built on rich metadata and seamless integration.

For example, the company’s forthcoming Intelligent Data Lake will offer zero copy integration and direct data catalog access to work with Oracle Database, Oracle HeatWave, Oracle Analytics Cloud, etc.—all using the same set of fine-grained, role-based access control policies.

Why is Oracle building this unified data layer, and what does it have to do with the company’s expanding OCI footprint across hyperscaler rivals? On the surface, Oracle wants to give customers a good reason to keep their data within the Oracle Database portfolio. Why pay to move and/or maintain vector data within an AWS or Google database when it is easier and cheaper to manage that information directly within Oracle Database 23ai? Moreover, why feed those vector embeddings to an external large language model (LLM) when it can be run directly within that same database at a lower cost and potentially lower latency?

These subtle efficiencies will mean the world to companies as they seek to scale up their AI investments. Oracle’s multi-model database architecture along with its in-database AI capabilities can radically simplify and streamline more complex GenAI data pipelines, even if those pipelines venture outside of the walls of the database itself. To that end, Oracle Autonomous Database Select AI can plug directly into LLMs provided by Microsoft Azure AI, Google Vertex AI, and Amazon Bedrock.

With this kind of advantageous adjacency in hand, Oracle can make a solid case not just for its database but for its entire OCI software and hardware stack. By making Oracle databases more appealing on external hyperscale platforms from Google, AWS, and Microsoft, the company headquartered in Austin, Texas, has created a sovereign state for Oracle Cloud itself, a bastion from which it can expand its lucrative suite of Oracle Fusion Cloud Applications, bringing those applications with integrated GenAI closer to customer data, no matter where it lives.

Appendix

Further reading

“The art of unlocking value through smart data fabrics – Lessons from InterSystems” (August 2024)

Market Landscape: Vector databases powering Generative AI (July 2024)

“Looking past the AI branding hype and toward increasingly unified GenAI platforms” (June 2024)

Author

Bradley Shimmin, Chief Analyst, AI Platforms

[email protected]