During AI World earlier this month, Oracle made a slew of AI announcements. This was to be expected, as the hyperscaler focuses on everything from the underlying hardware to the application layer. It is strongly promoting its own expansion of AI infrastructure. In addition, it is positioning itself as the most obvious system of record for AI.
Oracle certainly hopes to influence its customers’ thinking, but freedom of choice remains a priority. Multicloud Universal Credits are the best example of this. A new licensing option makes it possible to purchase Oracle AI Database and Oracle Cloud Infrastructure services on other clouds. This allows customers to choose dynamically between Oracle Database@AWS, @Azure, @Google Cloud, and OCI itself. It is a positive and logical next step for Oracle after it buried the hatchet with old rival AWS at last year’s CloudWorld (now rebranded to AI World). A consistent experience with Oracle’s products makes them more attractive to retain or switch to. After all, there should be as little friction as possible for customers considering a cloud migration. Anyone interested can request early access.
The AI expansion
Obviously, Oracle’s favorite customer is the kind to fully immerse themselves in OCI, Oracle AI Database, and Fusion applications. Oracle will tell you that they provide the aforementioned freedom of choice in the interest of the customer, and there is certainly an advantage to be had for that group. However, the hope on Oracle’s part is, of course, that one sees the benefits of a larger ecosystem adoption. Companies worldwide already entrust Oracle with mountains of data and organize it in its longstanding database product, now equipped with the latest update in the form of AI Database 26ai. As we wrote earlier, this is a drop-in replacement for 23ai. According to Oracle, it has become “AI-native” with support for agents, Apache Iceberg data formats, and stronger integrations for Databricks and Snowflake, among others.
The software story is one we will delve into further later. But the metal powering Oracle Cloud Infrastructure is considered the AI engine for OpenAI, among others, today. The extremely expensive AI chips make up a large part of the estimated $500 billion data centers from the Stargate project. Oracle, together with OpenAI and SoftBank, is attempting to build AI locations across North America; the Abilene site in Texas takes the crown with a predicted capacity of 1.4 gigawatts. There is also good news for European organizations, albeit with less impressive figures: around $1 billion will be invested in the Netherlands and $2 billion in Germany, for example.
More than the neoclouds
With such ‘exascale’ AI, the bottleneck is no longer the chip, Oracle argues. That is why it is focusing on the network with the newly announced Acceleron. This is a combination of solutions, from SmartNICs to software improvements. Acceleron’s magic word is ‘disintermediation’, or reducing the number of steps (hops) that network traffic has to take. This traffic is also more secure than ever thanks to Zero Trust Packet Routing (ZPR), which guarantees least-privilege access regardless of the exact network topology. In other words, you can expand your network however you want, and the data will remain secure. Acceleron promises to deliver three times faster IOPS for storage access and double the speed for packet processing. In addition, it would be cheaper than competing solutions.
When asked by Techzine, Oracle also commented on so-called ‘neoclouds’. Think of CoreWeave or Nebius, suppliers of AI infrastructure in the form of mountains of GPUs. They don’t do much else, which is why we wondered what Oracle actually thinks of these up-and-coming players. After all, CoreWeave is a new competitor of Oracle. Due to the insatiable appetite for AI computing power, such neoclouds are riding high. Yet we hear from Oracle that they have “a difficult business case” for the longer term. After all, they simply offer a lot of chips from a third party, without any software story or other differentiator that truly sets it apart. Oracle does offer that, such as a broad package of applications that can convert all AI infrastructure into AI applications, which is ultimately what it’s all about.
Agents dominate
The interoperability of Oracle’s systems makes them suitable as a foundation for AI adoption. AI Database 26ai and the Autonomous AI Lakehouse are an attractive combination for this, separate from OCI. With the new AI Data Platform, it packages this into a simplified experience “from raw data to production-grade AI.” Thanks to a collaboration with Nvidia and out-of-the-box support for agents with Agent2Agent and the Model Context Protocol, it complies with the conventions that have now been established in the GenAI landscape.
The AI agents and systems do not have to come from Oracle, but that is certainly an option. Within Oracle’s own software suite of Fusion applications, Ask Oracle reigns supreme, a universal assistant that effectively serves as an alternative to dashboard navigation. As we saw earlier with NetSuite, Ask Oracle is designed to eliminate the need to click through to relevant data. Not only that, it is designed to answer complex questions using criteria that you do not need to know explicitly. If you need to know which expenses are claimable according to company policy, you can simply ask Ask Oracle. It should provide the correct information using the knowledge base and other data in Fusion apps. The recommendations even appear as you type, thanks to ‘typeahead search’.
Within Fusion apps, Oracle offers various agents that automate common tasks. Think of purchasing departments that can compare products with the predictably named Product Comparison Advisor Agent, or management teams that require an exception for a single order via the Order Exception Assistant Agent. It’s clear that there is already a proliferation of agentic workflows for virtually all tasks in Oracle applications. Where an agent is missing, one can be built with the AI Agent Studio for Fusion Applications. And if an external agent is needed, MCP and A2A support ensure that it can be connected in a standardized way.
Conclusion: one whole becomes modular
According to analysts, Oracle has an “advantage in monetizing the AI platform shift thanks to its vertical integration of infrastructure, database, and apps.” That is certainly the case. But the most important detail here is that Oracle is always available to sell you something, whether you are on OCI, using an Oracle database, or using Fusion apps. The company makes it as easy as possible to adopt Oracle anywhere, so that it can always be available.
That vertical integration could be stronger. Oracle could have been a lot more sparing with external players. However, that would have made it less suitable as a starting point for AI. “Oracle databases are the system of record par excellence,” we hear from the company. It clearly does not want to squander that trust. After all, AI adoption could be such a major shift that you would be willing to endure painful migrations. In that case, you would choose IT players that integrate and do not lock you in. And, admittedly, migrating from Oracle can be painful, so existing customers will be quite motivated to stay. It seems that in 2025, the company will do everything it can to make that as attractive as possible.
Read also: Oracle reins in LLM reliance for maximum AI impact