Supercomputing and AI: Europe is making progress, and Italia is playing a leading role

But chips alone are not enough. We need computing infrastructure on a continental scale, and this is where EuroHPC comes in. The joint initiative has already rolled out a network of world-class supercomputers, and Europe has just entered the exascale era with the commissioning of Germany’s Jupiter, the first European system capable of exceeding the threshold of one billion billion operations per second.

Italia is not standing idly by: Leonardo, housed at Cineca, remains one of the most powerful supercomputers in the world, and with the Lisa upgrade it has gained specific capabilities for the training and inference of artificial intelligence models.

The race is also being played out on the private investment front. In May, a consortium comprising Leonardo, AI4Industry and Eni put Italia forward as a candidate to host a European AI Gigafactory – a computing facility dedicated exclusively to the training of large-scale artificial intelligence models. The proposal envisages the facility being built in Turin, where there is already a wealth of expertise in the aerospace and automotive sectors. “Italia has what it takes to become a leading European hub for artificial intelligence,” commented a spokesperson for the consortium. If the project goes ahead, it would add a crucial piece to the national strategy for digital sovereignty.

The issue of computational sovereignty is closely intertwined with that of security and strategic autonomy. Training AI models on national or European infrastructure means retaining control over the data used, the algorithms developed and the applications derived from them. This consideration applies equally to scientific research and to industrial and military applications. It is no coincidence that the National Cybersecurity Agency is an active partner in the EuroHPC projects hosted in Italia.

The challenges remain significant. Energy consumption in AI-dedicated data centres is growing at an exponential rate, raising questions about the environmental and economic sustainability of this race. The availability of specialist skills is another bottleneck: training engineers and researchers capable of making the most of this infrastructure requires time and investment in training.

And global competition waits for no one: whilst Europe is launching its first exascale systems, the United States and China are already planning the next generation.

And yet, the message emerging from Bologna is one of cautious optimism. Italia now has an ecosystem that combines traditional supercomputing, quantum computing and AI-dedicated infrastructure, all under European governance.

It is a starting point, not an end point, but it sets out a clear direction: to lay the foundations for technological autonomy that will enable Europe to compete in the new economy of artificial intelligence without being entirely dependent on powers outside the continent.