Neil Sholay is vice-president of Oracle, where he directs the initiatives linked to AI and the creation of commercial value. He returns for JDN on Oracle’s strategy in the field.
JDN. Oracle has been developing a strategy focused on AI agents for several months. What will be the decisive step to develop truly intelligent and autonomous agents, capable of acting without constant human supervision?
Neil Sholay. The agentics represents in my eyes the future of technological development, although I do not consider it as a final outcome. In the next 12 to 18 months, Oracle will gradually transform its business and SaaS applications by integrating agent workflows. These flows can mobilize up to two dozen individual interconnected agents, offering users a flexible customization and combination environment.
My deep conviction is that we are going to attend a transformation of intelligence, like the Cloud revolution a few years ago. Just as the Cloud has changed economic models by allowing to pay only what is used, intelligence will become a consumable service via an API. The history of technological innovations generally follows three phases: a first step of limited intelligence, a proliferation phase where market players build complex solutions, and finally the emergence of global platforms which naturally integrate these features.
From Oracle’s point of view, we are already starting from this third stage. Our huge base of SaaS and on -site customers allows us to deploy AI in an immediate and massive way. Even if we spent the next two years simply to activate AI for our existing customers, we would generate considerable value. Our goal is not to create AI for AI, but to make it accessible, useful and truly transformative for businesses. The autonomy of agents will gradually come, but always at the service of concrete business objectives.
To what extent could the emergence of the AC transform the very nature of AI agents, beyond simple operational autonomy?
If we collectively reach this level of intelligence, IA agents would become almost instantly obsolete, reduced to a minor element in a much larger technological paradigm. The emergence of AGA would radically move our questioning: we would no longer talk about the autonomy of agents, but the ability to apply truly general intelligence to solve complex problems in the real world. The open prospects would be revolutionary, as Larry Ellison pointed out during the announcement of Stargate.
Imagine the possibility of developing personalized vaccines in a few days, designed specifically for an individual. Or the detection of cancer in almost real time, with a precision and a hitherto unimaginable speed. These potential would largely transcend the current capacities of IA agents. In this new paradigm, current agents would become almost anecdotal. They would barely be mentioned, as the power of the act would make their current derisory features. It would no longer be a question of programming tasks or workflows, but understanding and solving problems on a scale and with a hitherto unprecedented depth.
AGE would thus represent a technological breakdown much more than a simple evolution. It would be a qualitative jump that would radically transform our relationship to artificial intelligence, going from automation tools to real cognitive partners capable of understanding, analyzing and solving complex problems with comparable autonomy and creativity – even superior – to those of humans.
Can you remind us of the precise involvement of Oracle in the Stargate project and give us an overview of its current progress?
The Stargate project is an ambitious initiative that brings together several major technological partners, including Oracle, Openai and SoftBank. Oracle plays a fundamental role in this project, taking responsibility for providing the technological infrastructure necessary for this new generation artificial intelligence adventure. The project is structured as a kind of joint venture or investment funds, with a financial commitment from each partner to build data centers in North America. Other technological actors such as NVIDIA and MCX are also expected to contribute to this initiative.
The first deployment is planned in Abilene, Texas, with the construction of data centers of a particularly advanced technological level. These infrastructure will integrate advanced technologies, including a liquid cooling system to the level of the chip and the use of renewable energies. The objective exceeds the simple construction of infrastructure. It is a question of bringing together a collective expertise to meet the challenges of the next border of artificial intelligence. By pooling the skills of four or five technological partners, the project aims to create a single synergy.
On the operational level, Stargate now exists as a legal entity. The next crucial step is to build the physical infrastructure, build teams, collect and structure data, and start building models. The first concrete achievements are expected during this calendar year, and not in several years as we could suppose
Your direct competitors like AWS and Google Cloud develop their own generative AI models. What motivated your decision not to follow this path?
Our strategic choice is based on the conviction that customers are looking for a diversity of language models. It is not realistic for a company to limit itself to a single large model. We anticipate that organizations will use several models, probably two, three, or even four. Moreover, we already observe the emergence of specialized models, designed for specific industries or segments. By offering this choice, we precisely meet the expectations of our customers.
Will Oracle never develop his own models?
No. My reflection is based on the observation of technological trends, which I have witnessed for almost 30 years in the industry. Generally, these technologies follow a similar scheme: a gross capacity emerges on the market, first offering basic use cases. Then comes a second crucial phase: a massive influx of start-ups and organizations develop tools, assets and services around this technology. This step offers more choices to customers, but also implies that they must integrate everything themselves.
Historically, these markets converge in two years to a complete end -to -end platform, where these capacities simply become an integral part of a main application. At Oracle, we anticipate that artificial intelligence will in the future accessible via an API, like the cloud, databases and storage. Its cost will balance and become accessible. Our strategy is based on clear choices: focus on our strengths rather than engaging in highly competitive market areas likely to quickly reach a tray. We will therefore focus on cloud infrastructure, IA infrastructure, the data layer that feeds models, preconceived AI services and associated applications.
Do you have new investments in Europe in your boxes, particularly in the fields of AI and Cloud?
In Europe, our investments focus on several strategic axes. We have already developed sovereign clouds in Frankfurt and Madrid to meet European regulatory requirements. In the United Kingdom, in connection with the recent announcement of investments of 14-15 billion pounds, we plan to contribute to the establishment of centers of excellence in Liverpool, Oxford and Cambridge. Our strategy is to position ourselves at the heart of national innovation ecosystems, by establishing close partnerships with European technological hubs. The objective is to offer an advanced cloud and AI infrastructure, sovereign and perfectly suited to local needs.
