State AI: chips, networks and energy decide who controls innovation

Until a few years ago, the cloud and cross-border data flows fed the illusion of a largely unified technological world. Generative AI broke that balance. Large language models, energy consumption, advanced microchips and data centers built at industrial scale made one thing clear: AI is "a vast and highly concentrated industry".

Who controls computing

That is the thesis of the analysis by the Institute for International Political Studies cited by CorCom. When computing capacity becomes scarce and expensive, whoever controls its distribution controls much of where innovation goes. Hence the choices made by the three players. The United States uses control of computational capacity as leverage. Europe has decided to build AI Gigafactories so it does not depend on outside infrastructure. China has folded the state and its companies together, and in some cases brings private operators directly into the governance of infrastructure.

The Commission's line

In her State of the Union address, European Commission President Ursula von der Leyen placed AI among economic infrastructures and security factors. She set out two parallel tracks: control of the most advanced technologies on one side, and on the other a European base of computing, funding and industrial applications. The document lists five priority application areas, among them health and advanced manufacturing, and calls for action on deepfakes and the protection of minors.

The thread running through it is that networks, chips and critical raw materials underpin technological sovereignty. Without them, rules stay statements of principle. So the contest is not only about models. It is about energy at competitive prices, ultra-broadband, and the capacity to produce components.

The consequences for companies

For Italian companies the message cuts both ways. Public policies put resources on the table for computing and for AI adoption. At the same time, the concentration of supply increases the bargaining power of the few suppliers able to offer infrastructure at scale. In this scenario, competitive differentiation depends less on the model chosen and more on proprietary data, on the quality of processes, and on the capacity to integrate AI into existing production systems.

"Digital sovereignty is not bought: it is built with energy, networks and silicon" is the conclusion shared by the two analyses.