Ivo Ivanov is the CEO of DE-CIX.
The concept of sovereignty has become synonymous with national borders. You can’t really have one without the other. Data sovereignty, however, isn’t about neat lines you can draw on a map. Enterprises are being propelled toward an AI future built on moving vast sums of data quickly, and that data may need to traverse the globe, or at least a continent or two. We talk about cities becoming “digital powerhouses”, but the next great AI hub may not be a city at all.
The reason boils down to basic geography. AI needs enormous quantities of power, land, compute and network connectivity, but the best place to secure one resource may not be the best place to find another. Training clusters, for instance, can be built where energy and space are readily available, but privacy, governance and sovereignty requirements may restrict where the data they use can travel. Likewise, inference increasingly has to happen near users, factories, vehicles, hospitals and other places that demand real-time AI. For certain AI workloads, forcing every part of that chain into a single metropolitan hub is impossible, so cross-border data travel has become an inevitable part of doing business.
That change is already underway. Data from DC Byte shows that “growth is shifting away from the largest and most congested” infrastructure markets, with AI accelerating development across emerging locations from the Southeast U.S. and Southern Europe to South and Southeast Asia. The approaching wave of real-time inference will spread that infrastructure even further. The geography of AI is therefore beginning to stretch beyond individual cities, distributing its parts across interconnected regions, with areas like the Iberian Peninsula in Southern Europe already offering an early glimpse of a future “AI hub.”
From Digital Capitals To Digital Regions
First, established digital capitals such as Frankfurt, London and Amsterdam aren’t going to disappear or lose relevance. Each city has spent decades accumulating the dense concentration of data centers, cloud platforms, network operators and Internet Exchanges (IXs) needed to keep digital services moving—and this will continue. However, pressure on local land, power and infrastructure is reaching a tipping point, and they can’t expand indefinitely. The average European data center site planned for 2026–2028 will be located 175 kilometers from a major hub, compared with just 46 kilometers for projects delivered between 2022 and 2025, as power and suitable land become harder to secure in cities such as London and Frankfurt. These hubs may become more like anchor points in a much larger regional system.
Take Iberia, for instance. Madrid, Barcelona and Lisbon are distinct digital markets in their own right, but they’ve become increasingly connected through a wider system of data centers, terrestrial networks, IXs and subsea cables. Together, they can combine access to power and new infrastructure capacity with connections extending across the continent and toward the Americas and Africa. No single city has to provide every component of the AI infrastructure chain because the region can distribute those components and connect them closely enough to function as one. In the coming years, this will begin to define an AI hub—several places with different strengths, held together by the data flowing between them.
How To Make A Moving Asset Like Data ‘Sovereign’
Data sovereignty is much simpler when data is idly stored. But we’re becoming increasingly dependent on data in motion. Take a modern smart car, for example. During a single journey, it might need to exchange performance data with its manufacturer and route data with its navigation provider, driver analytics with its insurer and endless infotainment and streaming service data with multiple providers. Those services will likely run across several clouds spanning several jurisdictions, depending on where the car is. While the driver has a single, seamless experience, the car’s data is moving through dozens of countries and regions.
It’s a good example because keeping all that data within one geographical area undermines many of the services the car is designed to provide. But who can access that data at each stage, and does the data owner retain full control once it passes into another jurisdiction? Those questions are already being asked about other data-fueled applications in healthcare, manufacturing, financial services and public infrastructure—where AI inference is becoming increasingly common. Distributed intelligence depends on data moving between organizations and locations, so data sovereignty can no longer be treated as a wall around a single location.
Sovereignty Has To Travel With Data
Data moves like water, flowing in and out of various regions and jurisdictions. Sovereignty therefore has to take on similar properties. Rules and controls must travel with information as it passes between devices, networks, clouds, partners and AI infrastructure, which requires targeted transparency rather than hard barriers. Organizations need to know where critical workloads can run, who can access them, which providers are involved and what happens when traffic crosses a regional boundary.
Interconnection is one mechanism that can make data movement more deliberate and controlled. Direct and private connections let organizations exchange critical traffic with clouds, partners and service providers without exposing it to the opaque, often congested public internet, while neutral IXs provide access to multiple networks without locking the organization into a single route or provider. None of this makes a workload sovereign on its own, but it does give businesses more visibility and control over data pathways, along with the freedom to change providers when operational, regulatory or geopolitical conditions shift. One mistake the industry has made is treating localization like isolation, but sovereignty needn’t mean building digital walls. Data is moving, and sovereignty needs to move with it.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?