Europe’s future as a leader in artificial intelligence depends on data centre infrastructure. That infrastructure will only be scaled if the continent rewires how governments, grid operators, manufacturers and hyperscalers work together.
On the edge of the Atlantic, where the fibre-optic cables from America come ashore in Portugal, a new model for AI infrastructure is taking shape. The Start Campus in Sines, a vast, purpose-built hyperscale site, is not just another data centre. It is a proving ground for Europe’s readiness to compete in the global AI economy. For Pablo Ruiz-Escribano, Senior Vice President for Secure Power & Data Centres, Europe at Schneider Electric, it symbolises how much is left to do.
“There is a huge opportunity in front of us,” Ruiz-Escribano says. “But we are running behind compared to the US. Other markets are much more flexible and able to let this happen faster. So the risk is that we miss this opportunity, which is why collaboration between public administrations, hyperscalers, co-location providers, utilities and manufacturers is crucial.”
Europe’s data centre capacity is expanding, but the surge mostly meets traditional cloud demand. AI’s rise has added a second wave of disruption that current systems were never designed to support. A fundamental redesign is needed, one that embraces the distinct physical and computational requirements of AI training and inference.
“The main difference is in the cooling,” he explains. “Storage and cloud computing require efficiency, but AI changes the game completely. The way we cool the data halls must change, because the temperature demands from AI chips are much higher. When latency requirements vary, training can be done further from demand, but inference needs to be local, just like online services. But the real challenge is that we have not had time to build for this. Now we are trying to catch up.”
Power is available but not accessible
From a distance, a data centre may look like a warehouse. But the comparison ends at the facade. Internally, an AI-capable facility’s tolerances, cleanliness, and precision rival those of a high-performance laboratory. And unlike anything that came before it, the energy scale required, particularly for GPU-rich AI training clusters, puts a strain on the surrounding grid.
“We do not lack power in Europe,” Ruiz-Escribano insists. “We have power. The challenge is access. Accessibility is the problem, not availability. We must rethink how the grid works with data centres. This is not just a technical issue, but a structural one. Grid investment timelines often stretch five years or more. Data centres, especially AI-focused builds, must be planned and deployed in under three. The mismatch puts Europe at a strategic disadvantage.
“If we design data centres with a three-year horizon, we cannot align with five-year grid investment cycles. So grid operators and data centre developers must share pipelines and plans to invest together, in the right places, at the right time. It is about aligning infrastructure development across sectors.”
From grid stabiliser to virtual power plant
The grid itself may have more to gain than most realise. Ruiz-Escribano believes data centres should be treated as grid assets, particularly in renewable-rich regions like Iberia, where solar and wind output often exceeds local demand. “Today, we are even curtailing renewable energy because the grid cannot absorb it all,” he says. “We can stabilise the grid by placing AI data centres where that energy is generated. In some scenarios, a data centre could act like a virtual power plant, dynamically balancing supply and demand.”
Portugal and Spain are not only capable of producing abundant green power but are also geographically privileged. Situated at the intersection of Europe, the Americas, Africa, and the Middle East, the Iberian Peninsula offers both low-carbon energy and global connectivity. This makes it a strategic hub for AI infrastructure, particularly with transatlantic fibre routes making landfall nearby.
“There is no other region in Europe with this density of fibre infrastructure,” he explains. “Add that to the fact that land is relatively cheap, the quality of life attracts global talent, and you are connected to South America, Africa, and North America. It is a unique combination. And when you have campuses like this one, Start Campus, designed from day one for sustainability, it becomes a model that others can replicate.”
Speed, sovereignty and strategic scale
Yet even ideal conditions cannot overcome one of the thorniest problems in Europe’s digital development: speed to deploy. Across much of the continent, planning and permitting delays stretch for years. That is a lethal lag for an industry chasing the breakneck evolution of AI hardware.
“It can take one to three years to get permits, depending on the country, city, and regulations,” Ruiz-Escribano says. “But technology is moving so fast that if it takes three years, your design might be obsolete by the time you build. We are not asking to reduce compliance, but we must find ways to accelerate approvals without compromising environmental or security standards.
“Europe also faces a balancing act between its commitment to data sovereignty and the scale advantages of cloud-based AI infrastructure. The two are not mutually exclusive but require careful alignment between sovereign compute environments and large-scale cloud backbones.
“We can leverage knowledge from large-scale cloud deployments to build localised, sovereign data centres that still benefit from those learnings. The challenge is adapting design, operation, and efficiency principles to meet data residency and regulatory needs without compromising performance.”
AI’s people problem
One of the most overlooked aspects of Europe’s AI ambitions may also be the most critical: people. Data centres are no longer anonymous facilities humming quietly in the background. As AI makes them more complex, energy-intensive, and vital to national and industrial competitiveness, the talent required to build and operate them must evolve in kind.
“This is a new kind of infrastructure,” Ruiz-Escribano says. “Ten years ago, nobody spoke about data centres in this way. Now they are strategic assets. You cannot treat them like warehouses or office buildings. These are precision environments; the skills needed to run them are highly specialised.
“Attracting and training that talent must become a strategic priority. It will also take a concerted effort from both industry and academia to avoid a skills bottleneck that leaves AI data centres unstaffed. We are already working with associations like Spain DC and Portugal DC to develop training plans with universities. We are also collaborating with companies like Stack, which is creating its own university to train graduates specifically for this industry. We contribute through demos, augmented reality tools, and hands-on training to ensure they are ready from day one.”
Regulation as a competitive asset
Europe’s regulatory environment is often cited as a barrier to innovation. However, when it comes to data centres, Ruiz-Escribano argues that they may be Europe’s strongest asset if they are leveraged correctly. “Usually, we see regulation as a burden, but it creates an opportunity in this case. Sustainability is a perfect example,” he says. “Our clients, Google, Microsoft and others, are making public pledges. European regulations force high standards, and the industry often goes beyond them. That gives us an edge.”
The ability to deliver high-efficiency, low-impact data centres under stringent regulations becomes a form of exportable expertise. What works in Portugal today could be the blueprint for tomorrow’s Asia or Latin America. “Start Campus is an example of what Europe can offer the world. These are not just compliant data centres. They are best-in-class, and that expertise was developed here. That is a competitive advantage we can build on.”
Winning the race with collaboration
At the core of it all, Ruiz-Escribano returns to a single idea: collaboration. Whether public authorities reduce permit timelines, grid operators align with infrastructure investors, or companies co-develop training academies, progress will only happen when interests are shared.
“We all need to recognise this as a once-in-a-generation opportunity and stop protecting our silos,” he concludes. “It is not about Portugal versus Spain, or one company versus another. Suppose we can build an ecosystem where everyone in France, Italy, and Iberia is aligned. In that case, we can attract the investment, shape the regulations, and design the infrastructure to lead in AI.”
Without that shift, Europe risks standing still while the rest of the world races ahead. But with it, Europe could become not only a leader in AI infrastructure but also a model for how it is built sustainably, scalably, and strategically.
