-min.jpg){kind=small}
First published in Portuguese in Expresso.
Stand on the Atlantic coast at Sines and the logic is immediately clear. Subsea cables beneath your feet carry information between Europe and the Americas. Offshore wind and Portuguese solar make up the grid that is among Europe's cleanest, and increasingly among its most competitively priced for large-scale users. Cold Atlantic water cools better than any mechanical system. This is not merely a data centre site. It is a natural home for the infrastructure that will power the next century of European prosperity.
For centuries, Europe's wealth has followed geographic logic, not national borders. Ports where currents favoured trade. Railways through the passes that connected markets. Farms where the soil allowed it. The question Europe now urgently faces is whether it will apply the same thinking to AI.
Today, Europe is broadly contemplating two models for AI compute. Yet, neither fully serves Europe’s long-term interests.
One says rent what you need from foreign providers and regulate on top. The other demands complete sovereignty — equivalent capacity in every member state regardless of energy economics, cost, or physics. The first turns countries into permanent tenants. The second creates inefficiency dressed up as independence.
However, there is a better model for European AI. It’s a hub and spoke and not a map of equal dots. It is a network with real topology – and Portugal is positioned to be one of the most important hubs.
Portugal, and Sines in particular, as an AI infrastructure hub is not a political choice. It is a geographical one. With cheap renewable power, natural cooling from the Atlantic, and subsea cables linking Europe to the Americas and Africa, there are structural advantages in Sines that cannot be replicated by political will alone.
The same logic applies to Spain. In the Nordics, ample hydroelectricity, natural cold, and deep fibre infrastructure have already drawn some of the world's largest data centre investments. Southeastern Europe and the Baltics are also poised to emerge as another key hub with growing renewable capacity, improving connectivity, and the political ambition to become serious players in European digital infrastructure.
These regions are not arbitrary selections. They are the natural homes for Europe's AI Gigafactories where the European Union’s AI ambitions can be physically realised at scale. The AI compute hubs for the rest of the Continent's existing financial and industrial hubs.
In fact, the EU is beginning to recognise this. Its upcoming tender for five AI Gigafactories, combined with an extensive EuroHPC network of supercomputing sites, sketches the early outline of exactly this architecture. That is the right instinct. But public subsidy can only anchor the model — the main lift will be private enterprise, deploying capital at the speed and scale that procurement cycles cannot match.
Around the gigafactory hubs, spokes will carry sovereign control and regulated capacity to every corner of the continent. Europe's telecommunications infrastructure is already embedded in every market, city, and industrial zone. They are the natural builders of this layer and will transform the existing infrastructure across Europe into a web of AI-capable nodes: national cloud environments, edge compute for hospitals, transportation systems and power grids, and defence and intelligence networks.
The reason why this hub and spoke model works is because not every workload needs the same thing. Sensitive data — patient records, financial systems, government workloads — stay on sovereign infrastructure, close to home. Everything else routes automatically to wherever compute is cheapest, greenest, and most available: Portugal when solar is peaking or the Nordics when hydroelectric capacity is high.
The principle is simple: centralise efficiency, decentralise control.
Europe has begun laying the groundwork — the EuroHPC access rules and the EU Data Act's portability provisions point in the right direction. But for this model to work at continental AI scale, those foundations must harden into binding access agreements that prevent any single country from denying compute service to others and open standards with real teeth so workloads can genuinely move around the continent with ease.
European governments must also create the conditions – faster and simplified licensing, grid access, long-term power agreements – that make it attractive for the private sector to build at their side, bringing local communities with them.
Portugal understands this and is emerging as one of the most compelling hubs in this network. The government has signalled their ambition, moved to accelerate grid connection timelines for large data centre projects, and has positioned Sines as a national strategic asset for digital infrastructure investment. That combination of political will and physical advantage is precisely what serious private capital looks for. Others are following and the model is taking shape.
In a world where AI compute shapes the trajectory of economic growth, Europe must build a system in which dependence is a choice, not a condition.
.png)
.png){kind=small}
