The AI rollout will grind to a halt if it finds no trust among its prospective adopters. Oracle, too, has noticed that technical discussions are no longer sufficient. Organizations are increasingly asking what a new IT solution means for their business operations. We talk to the company’s experts to find out how it approaches this changing customer demand in 2026.
We discuss the matter with Michiel van Vlimmeren, VP of Technology, and Marcel Giacomini, VP Sales Consulting at Oracle. Their company actually consists of two parts: applications and technology, with both gentlemen focusing on the latter. Technology here refers to Oracle’s hardware and infrastructure side. Whether companies purchase bare-metal servers, run entirely on Oracle Cloud Infrastructure (OCI), or sit somewhere in between, Van Vlimmeren and Giacomini are often the trusted advisors for these parties’ IT infrastructure.
Trust on top of technology
Oracle offers a diverse range of infrastructure solutions. This can be seen as a spectrum from bare-metal instances with no Oracle software installed at all to the complete abstraction layer of OCI, embracing the public cloud entirely. Clusters of thousands of bare-metal instances, each with up to 192 cores, 2.3TB of RAM, and up to one petabyte of block storage, can serve as the basis for an AI infrastructure. The absence of Oracle software in these instances means that, in principle, you could migrate to another supplier without losing functionality. However, as mentioned above, the Oracle portfolio is a spectrum: the next step is to move to the cloud, but one can do so in a manner that holds sovereignty in high regard.
We ask the Oracle experts what sovereignty means for the company. After all, opinions on this topic are divided and often entrenched. Van Vlimmeren gives a clear answer that we have admittedly heard a fair few times before. According to him, sovereignty is about ownership of and access to one’s data. In his view, the concept is essentially a mixture of agreed data residency with access controls that shield the data from outside parties. Oracle itself suggests that there is no additional cost (“at the same price point” as the regular public cloud, we read here). That may be true for OCI itself, but we understand from customers that, in practical terms, there is certainly an additional cost to a sovereign setup. This is due to factors such as extra compliance measures and a limited supply of technicians (who, for example, may only be European or of a certain nationality).
The bare-metal cloud provider setup also allows Oracle to be used as the foundation for their own “sovereign” solution. This is called Oracle Alloy, which can run all kinds of variations of Oracle infrastructure, not just bare-metal instances. We put ‘sovereign’ in brackets with Alloy, not because we are skeptical about the control mechanisms, but because a Dutch or European cloud provider still uses Oracle’s offering behind the scenes in this way. The hardware can be located in a proprietary data center or run in an OCI region. Again, in such cases, the encryption may reside with the provider or even the end customer. Van Vlimmeren argues that such measures contradict “the presumption of American intervention” in the case of an American provider.
Localization
We see a partially logical focus on North America at Oracle. The lion’s share of what is called the Stargate project is seeing the light of day on that continent. Several tech companies, including Oracle, are investing hundreds of billions of dollars in it. The project involves a huge series of AI complexes that are used to build the latest OpenAI models and perform massive AI calculations. “We’re going to use [Stargate] to solve the most advanced problems in the world,” says Giacomini. It’s a vision of Oracle CTO and co-founder Larry Ellison who, as Giacomini describes it, wants to use the computing power “to tackle a whole host of problems.” In short, if that hope becomes reality, the whole world will benefit. Incidentally, Oracle is also investing a billion in Dutch cloud infrastructure, to name one local example, so the focus is not as narrow as some headlines would have one believe.
Such hyperscale initiatieves are admittedly far removed from the mostly limited IT requirements of local customers. The reason we are including it is that Oracle also “needs to be able to substantiate its proposition with local or regional examples,” according to Van Vlimmeren. Stargate alone can’t be more than a flagship, more concrete use cases need to exist as well. This involves making IT solutions tangible, from AI infrastructure to everyday servers for current business processes.
There is certainly a local Dutch example of this at Oracle’s AI World Tour in Amsterdam, in the form of Dienst ICT Uitvoering, or DICTU for short. We also spoke with Richard Wiersema, Director of Operations at DICTU. His Dutch government agency, which is responsible for the IT operations of several ministries and other government bodies, will launch an “agentic AI platform” at the end of this year. Although discussions with Oracle are at an advanced stage, Wiersema tells us that they are still working on a deal – and the eventual architecture. Nevertheless, there are already examples of AI being used in reality by the government. For example, AI checks whether photographic reports of earthquake damage in the northern province of Groningen are new and whether there is illegal fishing in a particular area. It saves manpower by using drones to take images and analyzing them with AI, instead of having inspectors visit the site.
Data, data, data
This is just a selection of what is possible with the right data and AI solution, which, through Oracle, should be able to provide the desired sovereignty, scale, and flexibility. Nevertheless, it remains a technical undertaking. Why doesn’t Oracle itself provide more ready-made tooling? In a later article, we will delve into the idea behind the company’s AI Data Platform, which serves more the business side of AI insights based on Oracle’s systems. But to get to that point, it is first necessary to have the technical foundation in place.
Oracle will never build a Large Language Model (LLM) itself, Giacomini says. Van Vlimmeren emphasizes that there is certainly a role for his company in educating customers about AI models and their usefulness. IT systems are currently focused on people, which AI agents sometimes seem to bypass in awkward ways. Just consider the fact that an AI agent sometimes still has to use a normal account for identification, when it would be much more logical to roll out a separate category for such AI systems. The data is also still hidden in silos or simply not available digitally. Modernizing all of this is possible, but it requires a pull factor. That could be AI, but Van Vlimmeren and Giacomini argue that there is still work to be done in that area. “Quick fixes,” as the former characterizes them, are not enough.
What that next step will be remains to be seen. As a trusted advisor, Oracle will continue to engage in discussions with customers who want to utilize AI. The step before technical integration revolves around confidence in compliance with sovereignty, the lack of lock-in, and other such issues. The integration itself is demanding. But then comes the change in business processes brought about by AI, a task that will probably take many years.