Trends

Supercomputers and agents, artificial intelligence becomes mature

According to Capgemini and Gartner, this year Ai will move from projects to actual entry into the enterprise

by Gianni Rusconi

16 January 2026

3' min read

Translated by AI

Versione italiana

Key points

The year of truth for artificial intelligence
Supercomputer for Ai
In field agents
Anti-Cybersecurity

3' min read

Translated by AI

Versione italiana

The year that has just begun is shaping up to be one of discontinuity for companies' technological strategies. After a phase dominated by experimentation, the focus is shifting to the ability (of innovation) to generate concrete and measurable value. AI, cloud and new architectures are configured as strategic choices that impact governance, skills and business models, and it is in this perspective that industry analysts have identified the trends that can guide investments in the coming years.

The year of truth for artificial intelligence

According to Capgemini, 2026 will first and foremost be 'the year of truth' for artificial intelligence. Proof-of-concept, isolated use cases and pilot projects (often fragmented and in several cases doomed to failure) will give way to a technology maturity phase, characterised by implementations in which AI becomes an integral part of an organisation's core architecture and processes.

We are therefore entering a season of 'proof-of-impact', where the value of projects will lie not so much in the technology itself as in the approach, the quality of the data and the ability to build a solid collaboration between people and intelligent systems. This evolution will also affect software, with AI and Llm models rewriting (speeding up and refining) the application development cycle under human supervision.

Supercomputer for Ai

At the same time, Cloud 3.0 will take shape and substance, i.e. a set of public, private, hybrid, multi-cloud, edge and sovereign architectures, indispensable to support large-scale artificial intelligence workloads.

Gartner experts, looking at the technologies set to reshape the next five years, describe 2026 as a 'pivotal year', marked by an unprecedented acceleration of innovation, risk and complexity. At the centre of the scene are supercomputing platforms for artificial intelligence, which combine Cpu, Gpu and new computing paradigms to handle extremely intensive workloads.

In the field specialised agents

According to estimates, by 2028 more than 40 per cent of large enterprises will adopt hybrid computing architectures in critical processes, with applications ranging from financial market simulation to energy network management. A key role, also highlighted by Capgemini, will be played by multi-agent systems, sets of specialised AI agents that collaborate with each other to automate and orchestrate complex 'end-to-end' processes and create new ways of collaboration between humans and algorithmic models.

In this scenario, the importance of so-called Domain-Specific Language Models (Dslm), trained on industry- or function-specific data and contexts to ensure greater accuracy, lower costs and better compliance than generalist models, will grow substantially. Gartner predicts that by 2028, more than half of the Gen AI tools used in enterprises will be of this type in order to meet the growing need for reliable results produced by machines.

Anticipatory Cybersecurity

Security will once again be a priority. AI security platforms will become essential to protect applications and intelligent agents from specific risks such as 'data leakage' or uncontrolled behaviour, while cybersecurity will evolve to 'preemptive' status, anticipating and neutralising threats before they occur through AI-driven solutions. Within the next four years, these solutions will account for about half of global security spending, while at the same time 'digital provenance', i.e. the ability to verify the origin and integrity of software, data and content, will become increasingly important.

More concrete trends include Physical AI, an already well-known paradigm that brings artificial intelligence into the physical world (industrial, energy and logistics environments) through robots, drones and 'thinking' machines capable of perceiving, deciding and acting: new skills integrating IT, engineering and operations will be needed to manage it.

Finally, there is one last strand to consider, which is mirrored in geopolitics: 'confidential computing' to protect data even while it is being processed and 'geopatriation' to move data and applications to local or sovereign solutions will be part of the strategies of more than 75 per cent of EMEA companies by 2030. The 'geopatriation' of data and applications will be part of the strategies of more than 75 per cent of EMEA companies by 2030