The Infrastructure Variable Nobody Talks About Enough
A decade ago, the conversation around European data centre competitiveness centred on tax incentives, fibre density, and the availability of technical talent. Government ministers published roadmaps. Development agencies structured grants. Planning departments streamlined approvals for campuses that would bring jobs and foreign investment. All of that still matters, but none of it is the decisive variable anymore. The variable that now determines whether a hyperscaler breaks ground in Dublin, redirects capital to Stockholm, or commits a multi-billion-euro programme to Madrid is something far less glamorous than a policy package: it is the lead time to secure a high-voltage grid connection, and how many other applicants are already ahead in that queue. Grid connection lead times across Europe’s primary data centre markets now range from three to ten years depending on the market, and that single fact is quietly reshaping the continent’s digital geography more decisively than any incentive programme ever could. Countries that invested in grid modernisation, transmission infrastructure, and substation capacity five years ago now receive disproportionate inflows of hyperscaler capital. Countries that did not are watching projects stall in planning and connection queues while their neighbours announce campus after campus.
The scale of demand driving these queues is unlike anything European grid operators have managed before. Data centres already consumed a significant share of electricity in Europe’s largest markets by 2024, with Dublin reaching approximately 80% of local electricity demand attributed to data centres, while Amsterdam, London, and Frankfurt operate in a range of roughly one-third to over 40% of local demand. These are not abstract system-level pressures — they represent physical, localised stress on infrastructure built decades ago for fundamentally different load profiles. Hyperscale deployments and AI clusters arrive as continuous, high-density loads, often exceeding 100 MW per application, with no ability to stagger or distribute workload without performance degradation. The FLAP-D markets Frankfurt, London, Amsterdam, Paris, and Dublin — grew from 1.8 GW of combined live capacity in 2019 to 3.6 GW by 2025, more than doubling in six years, and that doubling happened on grid infrastructure that was never designed to carry it. The connection queues that now stretch across Europe are the physical consequence of that mismatch between demand velocity and infrastructure investment cycles.
Why the Queue Is the Decision
The architecture of investment decisions in the hyperscaler world is not built on aspiration it is built on delivery certainty. When Microsoft, Google, Amazon Web Services, or Meta evaluates a European market for a new campus, the primary criterion is not the cost of land or the generosity of tax treatment. It is time-to-power: how quickly can a credible, firm, high-capacity grid connection be secured? JLL’s 2026 Market Outlook for Global Data Centers confirmed this shift explicitly, identifying speed to power as the primary criteria driving site selection decisions globally, ahead of community support, latency, and proximity to customers. That reordering has profound consequences for European markets because time-to-power is not a function of permitting ambition or political will it is a function of physical infrastructure that takes years, sometimes decades, to build. A substation upgrade requires civil engineering, equipment lead times, land acquisition, and operational commissioning. A new transmission line requires route surveys, environmental assessments, stakeholder consultations, and construction windows that cannot be compressed by legislation. When connection queues in major hubs stretch to ten years, no regulatory fast-track can substitute for the physical absence of spare capacity. Countries that understand this distinction are now making very different infrastructure decisions than countries that still believe policy declarations can substitute for civil engineering.
The geography of constraint is not evenly distributed across Europe. ENTSO-E, the European Network of Transmission System Operators for Electricity, has reached the same conclusion from a system-operator perspective: connection lead times can range from a few years to more than a decade in constrained regions, creating a time-to-power gap that is now central to data centre development strategy. Within that range, individual markets vary dramatically based on the investment decisions their grid operators made years before AI-driven demand arrived. Markets that invested early in transmission capacity, substation density, and grid modernisation now hold a structural advantage that cannot be replicated quickly — and the hyperscaler capital flowing into those markets is compounding that advantage with each new campus announcement. The divide between constrained and unconstrained European markets is not widening gradually; it is widening at the pace of AI infrastructure investment, which means it is widening fast.
What Conditional Access Actually Means for Dublin’s Pipeline
The CRU’s December 2025 decision is more sophisticated than the blunt instrument it replaced, but it does not resolve the underlying grid constraint. It restructures the terms on which new demand can access constrained infrastructure. EirGrid and ESB Networks received the authority to reject proposed sites for data centres that fail to meet energy supply requirements, and system operators will now publish enhanced, high-granularity network capacity information to enable more precise location decisions. The new policy applies only to applications lodged after December 2025 — existing applications continue under the 2021 framework, preserving a two-tier landscape that will take years to resolve. For operators with applications already in progress, the path remains governed by criteria written for a different era of grid stress. For new entrants, the conditional access model creates a viable but demanding pathway that shifts the cost and complexity of grid stability from the public system to the private developer. That shift is financially significant: onsite dispatchable generation at the scale required for a hyperscale facility represents substantial capital expenditure that must be factored into investment models alongside land, construction, and operational costs.
EirGrid’s large demand facility requirements, combined with the CRU’s renewable integration obligations, mean that Dublin’s recovery as a development hub will be measured in years rather than quarters. The policy framework now exists, but the physical grid constraints that triggered the moratorium have not disappeared — they have simply been reconfigured as requirements for the operator rather than barriers from the regulator. Spencer Lamb of Kao Data has characterised the FLAP-D markets as likely to retain latency-sensitive enterprise and inference workloads that require proximity to end users, even as large-scale AI training deployments shift toward regions with more abundant and cost-effective renewable energy. Dublin fits that description precisely: it retains strategic value for specific workload types, but its era as an unconstrained development market for any type of data centre has closed. The Large Energy User Action Plan, described as the framework enabling additional growth from 2026 onward, reflects a city that has accepted a more structured and conditional relationship with the sector that transformed its grid.
Amsterdam: The Market That Chose to Slow Down
Noord-Holland’s Regulatory Architecture
Amsterdam’s relationship with data centre development has been characterised by a series of increasingly firm interventions, each one pushing the market further from the open-access model that made the Netherlands one of Europe’s most important digital infrastructure hubs in the 2010s. The Amsterdam Internet Exchange remains one of the world’s largest internet exchanges, giving the city an inherent connectivity advantage that no regulatory framework can fully neutralise. That advantage has kept Amsterdam in the development conversation even as the restrictions have mounted. However, the restrictions have been structural, cumulative, and deliberate in a way that distinguishes Amsterdam from markets where constraints arose from grid stress alone. The province of Noord-Holland was the first regional government in Europe to draw up its own data centre strategy, adopted by the Provincial Council in January 2022, applying to new data centres with an area of more than 2,000 m² and an electricity connection of at least 5 MVA. That regional strategy created a planning architecture that no individual municipality could override, establishing geographic boundaries and performance requirements that redefined where development could occur.
The Dutch national government went further. An amendment to the decree on rules for the quality of the physical environment, which came into force on 1 January 2024, effectively prohibited hyperscale data centres across the Netherlands, with limited exceptions in designated locations including Eemshaven and Agriport A7 in Hollands Kroon. A hyperscale data centre under this definition means a facility of more than 10 hectares with an electrical connected load of 70 MW. The Amsterdam city government moved in parallel: in April 2025, the municipality announced it would allow no more data centres or expansions within city limits until at least 2030, with the exception of projects already in the pipeline. Alderman Zita Pels framed the decision in direct terms, stating that the city needed to reserve grid capacity for other uses. The Dutch Data Center Association responded with significant disappointment, noting that the moratorium caused international demand, particularly for wholesale data centre space, to drop and move to other European locations. Growth in the Dutch colocation market slowed markedly in the years following the first moratorium, with Amsterdam lagging its FLAP-D rivals in new capacity additions.
What Remains, What Redirected
The Amsterdam restrictions did not eliminate all development activity they channelled it. Projects with pre-existing approvals and legacy pipeline positions continued under transitional rules. Microsoft’s large campus development in Amsterdam proceeded because the initial plans predated the bans by nearly a decade, with the province of Noord-Holland approving separate permits for each tower under an interpretation of transitional provisions. That decision attracted controversy precisely because it illustrated the gap between the spirit of the moratorium and the legal mechanics of transitional planning law. Rick Pijpers, former director of Equinix and founder of the Dutch Sovereign Datacenter Cooperative, warned publicly that US technology companies were crowding out Dutch and European providers, with grid capacity becoming scarce for local operators. The Amsterdam market’s contraction did not create a vacuum it created a redirect. Demand that would historically have gone to Amsterdam’s established colocation ecosystem moved instead to Frankfurt, London, and increasingly to emerging markets that offered grid access without the regulatory complexity. Amsterdam’s market share within the FLAP-D cluster declined as a consequence, with the Dutch capital losing its position as the primary alternative to London for European connectivity-critical deployments.
The lasting significance of Amsterdam’s trajectory is not the volume of capacity it lost it is the signal it sent. A market that chose to restrict development for grid and land-use reasons demonstrated that European cities with high data centre concentration faced a genuine governance dilemma: the infrastructure that made them attractive also threatened to make them unmanageable. That dilemma is not unique to Amsterdam. It runs through every constrained European market, and the way different jurisdictions navigate it is now one of the most consequential variables in the continental distribution of AI infrastructure investment. Amsterdam chose restriction. Dublin chose conditional access. Other markets chose to compete on the basis of what they could offer rather than what they needed to limit.
Spain: The Renewable Dividend Meets Grid Reality
Why Madrid Is Europe’s Fastest-Growing Data Centre Market
Spain’s emergence as one of Europe’s most significant data centre markets is not accidental — it is the product of a specific combination of grid characteristics that took years to develop and that constrained markets cannot replicate quickly. The country holds one of Europe’s largest renewable power portfolios, primarily solar and wind, which means data centre operators seeking long-term power purchase agreements for 100% renewable supply can sign credible, large-scale contracts at competitive prices without the structural scarcity that characterises PPAs in constrained grids. Spain’s geography places it at the convergence of transatlantic and Mediterranean submarine cable systems, giving it a connectivity profile that supports both latency-sensitive and large-scale training workloads. The government’s 2023-2030 National Energy and Climate Plan positioned Spain as a frontrunner in the energy transition, and the regulatory environment for renewable development has been more permissive and more productive than the European average. Daniel Thorpe of JLL’s EMEA data centre research team identified Spain and Italy as Europe’s strongest growth zones, specifically noting that Madrid and Milan have been the fastest-growing markets from a percentage growth perspective a distinction driven primarily by the combination of favourable infrastructure, government incentives, and new subsea cable routes.
The Madrid market has attracted hyperscaler investment at a pace that reflects genuine confidence in grid access rather than opportunistic pricing. Amazon, Microsoft, and Google collectively have more than 1 GW under development across Spain. Iberdrola created a joint venture, Echelon Iberdrola Digital Infra, representing the largest partnership between an energy company and a data centre developer in Europe, with a portfolio of more than 700 MW secured and a potential pipeline of 5,000 MW. The first project, Madrid Sur, is a 160,000 m² complex with 144 MW for data processing that already has a 230 MW electricity connection secured a detail that would be unremarkable in a market with abundant grid capacity but is, in the current European context, a significant competitive differentiator. That connection certainty is what makes the project fundable and executable on a timeline that hyperscalers can commit to. Spain’s Ministry for the Ecological Transition has been finalising the 2025-2030 Electricity Grid Development Plan, which allocates additional capacity specifically for data centres, signalling that grid investment is being aligned with digital infrastructure demand at the national planning level.
The Grid Investment Gap Spain Still Has to Close
Spain’s competitive position is genuine, but it is not without constraint. The key bottleneck in the Spanish market lies not in generation capacity — of which there is abundant renewable supply but in securing timely grid access and connection, which remains hindered by slow permitting processes and limitations on network investment planning. A planned €13.59 billion grid investment by 2030 aims to improve grid access for high-demand users including data centres, and early site selection near grid upgrade zones will provide developers with a strategic advantage. The Spanish energy sector has seen significant interest from both domestic utilities and international capital, but the pipeline of proposed projects significantly exceeds realistic deployment capacity. Approximately 12 GW of capacity has been granted to projects linked to digital infrastructure, far exceeding industry estimates of 2 to 3 GW of deployable capacity by 2030. That gap between granted capacity and executable projects reflects a permitting and grid connection process that has not yet been streamlined to match the velocity of demand. The solar-plus-storage PPA model that has enabled sustainable campus-scale builds in Spain and Portugal provides a structural funding mechanism for behind-the-meter power that reduces grid dependency, but it does not eliminate it.
The distinction between Spain’s favourable trajectory and its unresolved constraints is important because it illustrates a broader truth about Europe’s data centre geography: no market is simply open or simply closed. Every market sits on a spectrum of accessibility determined by the interaction between grid capacity, renewable supply, permitting velocity, and regulatory clarity. Spain’s position on that spectrum is meaningfully better than Dublin’s or Amsterdam’s in the current cycle, but it is not unlimited. The markets that will capture the largest share of long-term AI infrastructure investment are those that address the physical infrastructure constraints proactively rather than reactively and Spain’s €13.59 billion grid investment commitment, if executed, is precisely that kind of proactive positioning. The challenge is execution timing: grid investment programmes measured in billions and designed for delivery by 2030 move on civil engineering timelines, not on hyperscaler announcement timelines.
The Nordic Advantage: Infrastructure Built Before the Rush
Why Norway, Sweden, and Finland Hold Structural Leverage
The Nordic countries occupy a fundamentally different position in the European grid queue conversation, because their competitive advantage was not designed as a response to data centre demand — it was an existing characteristic of their electricity systems that data centre demand found. Norway, Sweden, and Finland operate on grids powered predominantly by hydropower and, increasingly, wind, delivering grid stability characteristics and carbon intensity profiles that no thermal-heavy grid can match. The total cost of ownership for power-intensive AI workloads in Norway and Sweden is demonstrably the lowest in Europe, not primarily because of energy price arbitrage but because of the combination of ambient temperature-assisted cooling, renewable grid power, and grid stability that reduces both operational expenditure and infrastructure capital requirements. These are not advantages that can be legislated into existence in a constrained market. They are the product of decades of infrastructure investment and geographic fortune.
Microsoft’s commitment to Nordic data centre infrastructure runs to billions across Norway, Sweden, Denmark, and Finland. The partnership with Nscale for AI infrastructure in Norway, with an investment reported at approximately $6.2 billion, represents one of the largest single AI infrastructure commitments in Europe. A $3.2 billion expansion in Sweden and a $3 billion investment in Danish data centre capacity between 2023 and 2027 reflect a sustained, multi-market Nordic strategy that is driven by grid credibility rather than any single incentive package. AtNorth surpassed 200 MW of operational capacity across Iceland, Sweden, and Denmark in 2025 and announced a further 300 MW expansion pipeline. Nscale operates facilities in Iceland running on 100% renewable energy from geothermal and hydropower sources, with GPU installations at Verne’s Icelandic campus representing some of the largest liquid-cooled deployments in Europe. Nvidia’s own analysis of the European market has identified Norway as the leading destination for AI infrastructure investment, with Denmark and Sweden closely following — a ranking derived directly from power cost, grid reliability, and renewable supply certainty rather than from historical data centre market scale.
The Nordic Growth Trajectory and Its Limits
The Nordic markets are not immune to the pressures of rapid scaling. Grid-adjacent fibre infrastructure, long-haul interconnect capacity, and talent availability in smaller Nordic cities all require sustained investment as demand scales from research and backup workloads to active AI training at hyperscale. The Brookfield AI Data Centre in Sweden, designed at 750 MW of IT load with a footprint of 350,000 square metres and a projected 10 to 15 year build timeline, reflects the scale of ambition entering the Nordic market — and the long physical execution cycles that even favourable markets require. As Ember’s analysis of European data centre grid dynamics noted, the Nordics and Southern Europe, with uncongested grids and shorter connection wait times, are expected to see data centre demand grow at nearly double the rate of the traditional FLAP-D leaders. That projection carries an implicit warning: markets that are currently unconstrained will face their own connection pressure as investment accelerates. The window of advantage the Nordics currently hold is real but not indefinite, and the decisions those markets make now about grid investment, planning frameworks, and capacity allocation will determine how long the advantage holds.
The IEA’s analysis of European data centre energy constraints, published in November 2025, noted that Dublin and Amsterdam paused new projects due to grid availability limitations while Copenhagen and Milan began playing increasingly significant roles in European capacity. That shift is not accidental. It reflects a deliberate process of capital finding the path of least electrical resistance — and doing so faster than any regulatory framework can track. The Nordic markets benefit from a grid history that made them natural recipients of that redirected capital. The question for Nordic grid operators is not whether demand will arrive — it has arrived — but whether investment in interconnection, substation expansion, and distribution infrastructure will keep pace with the velocity of hyperscaler commitment. Markets that have answered that question positively, with funded grid expansion programmes and transparent connection timelines, will retain their advantage. Markets that allow queue formation to develop will find, as Dublin and Amsterdam did, that the capital simply moves on.
What the European Commission’s Tech Sovereignty Package Can and Cannot Do
CADA and the Infrastructure Gap
The European Commission published the European Technological Sovereignty Package on 3 June 2026, representing the EU’s most ambitious assertion of digital autonomy to date. The package combines two legislative proposals — the Cloud and AI Development Act and Chips Act 2.0 — with an EU Open Source Strategy and a Strategic Roadmap for Digitalisation and AI in Energy. CADA, the centrepiece from a data centre infrastructure perspective, creates a framework for Cloud and AI Leadership Initiatives designed to streamline permitting, integrate data centres with electricity systems, and create designated data centre acceleration zones. The proposal aims to triple EU data centre capacity over the next five to seven years, ensuring the Union has the capacity it needs by 2035. Commission President Ursula von der Leyen framed the package in terms of strategic dependency: Europe cannot afford to depend on others for the technologies that keep its hospitals running, its energy grids stable, and its services secure. That framing reflects a genuine policy concern about the concentration of critical digital infrastructure in non-European hands, and CADA’s sovereignty assurance framework for public-sector cloud workloads addresses a real jurisdictional exposure that became impossible to ignore after testimony before the French Senate in June 2025 clarified the limits of contractual protections from US cloud providers.
The STL Partners analysis of the EU AI Gigafactory Initiative reached the same conclusion from an infrastructure perspective: AI gigafactories need 200 to 500 MW of contracted, low-carbon power and a permitting pathway that can deliver capacity in three to five years. That combination rules out FLAP-D metros and directs the build-out toward Member States with grid headroom, hydropower, or nuclear capacity. CADA’s data centre acceleration zones are designed to create precisely that kind of fast-track permitting pathway. Operators with land banks, substation relationships, and grid-adjacent fibre in favourable geographies will anchor the consortia that win sites. The EU’s InvestAI initiative, launched in 2025, seeks to mobilise €200 billion for AI investments, and the AI Continent Action Plan targets at least tripling data centre capacity within five to seven years. The legislative ambition is clear. The physical constraint is also clear. No amount of streamlined permitting can compress the time required to build a substation, upgrade a transmission line, or commission a new generation source. CADA creates conditions for faster development where grid capacity already exists — it does not create grid capacity where it does not.
The Permitting Gap Versus the Civil Engineering Gap
The distinction between permitting timelines and civil engineering timelines is not a technocratic footnote it is the central analytical problem with European data centre policy. Permitting reform can compress a two-year approval process to six months. Civil engineering cannot compress a seven-year substation programme to two years. European policymakers have a tendency to frame grid connection delays as regulatory problems, solvable through directive and coordination. Some of the delay is regulatory: inconsistent national permitting frameworks, slow environmental assessment processes, and poor inter-agency coordination all add time to connection timelines that proper reform could reduce. Akin Gump’s analysis of CADA notes that the legislation integrates data centres with electricity systems and creates designated acceleration zones both of which address genuine regulatory friction. The Strategic Roadmap for Digitalisation and AI in Energy, also part of the package, includes a rating scheme for data centre energy performance to be adopted in 2026 with first labels in 2027, and a needs assessment for minimum EU energy performance standards due by 2027. These are important policy instruments for aligning digital infrastructure with the energy transition.
What CADA cannot do is replace the physical investment cycles that determine actual grid capacity. The lead times that ENTSO-E documents stretching beyond a decade in the most constrained regions exist because transformers have long manufacturing lead times, transmission corridors require land acquisition, and grid upgrades require outage windows that cannot be scheduled around market demand. Ember’s analysis of grid strategy for European data centres identifies ambitious grid planning as the mechanism that can accelerate deployment and the operative word is planning, not legislation. The countries that are winning the European AI infrastructure race today made their grid investment decisions in 2019 and 2020, when the AI demand wave was not yet visible on any grid operator’s forecast model. By 2026, their early investment has translated into connection certainty that CADA cannot legislate into markets that made different decisions. The package is a meaningful and necessary step, but its impact on the current investment cycle will be modest. Its impact on the cycle that follows — the post-2030 wave depends entirely on whether Member States use the legislative and financial frameworks it provides to fund physical grid infrastructure now rather than waiting for the next demand crisis to force the decision.
The Secondary Markets and the New Geography of European Compute
Where the Overflow Is Going
The saturation of primary markets has not simply reduced European data centre investment — it has redistributed it. The FLAP-D cluster remains Europe’s most mature data centre group, but its relative weight is declining as power constraints, permitting complexity, and land scarcity push new developments outward. Rabobank’s 2026 research on European data centre geography documents this shift with precision: new capacity builds are increasingly concentrated in Nordic markets, Spain, Belgium, Italy, Poland, and Portugal. These markets benefit from improved connectivity, renewable power availability, and more favourable development conditions than the saturated primary hubs. Italy’s Milan market, supported by expanding subsea cable connectivity and strong renewable integration, has emerged as one of Europe’s fastest-growing data centre locations. Poland has attracted investment due to its grid expansion ambitions and its strategic position connecting Central and Eastern European digital infrastructure to Western European consumption centres. Belgium, with its dense network of interconnection points and improving grid investment programme, has positioned itself as a credible secondary market for operators seeking proximity to major European business centres without the connection constraints of Amsterdam or London.
The market geography that emerges from constraint is not random it follows the physical logic of grid availability and renewable supply in exactly the same way that water flows to low ground. Secondary markets benefitting from the redirection of primary market demand share a common characteristic: they invested in the conditions for data centre development before that development arrived in volume. Portugal’s Lisbon market offers grid access without the constraints of major metropolitan areas. Manchester provides UK-based operators with an alternative to London’s compressed connection timelines. The JLL EMEA year-end data centre report for 2025 documented that over half of AI growth is expected in Tier 2 and emerging markets precisely because grid lead times of up to ten years have redirected demand away from primary hubs. That projection is not a forecast of what might happen — it is a description of what is already happening, visible in the campus announcements, hyperscaler leases, and development pipeline data accumulating across markets that were considered secondary as recently as 2022 but are now receiving capital that would previously have gone to Dublin, Amsterdam, or a third Frankfurt substation.
The Structural Divergence and What Drives It
The bifurcation in European data centre development — constrained primary markets on one side, unconstrained secondary and emerging markets on the other — will not resolve itself when CADA is implemented or when individual markets reform their permitting frameworks. It will resolve itself when the physical grid infrastructure gaps close, and those gaps close on civil engineering timelines measured in years and decades. The divergence has been structural from the beginning: it reflects decisions about electricity grid investment made at the national level years before AI demand made those decisions consequential. What changes now, with the availability of EU legislative frameworks, InvestAI capital, and CADA’s acceleration zone architecture, is the visibility of those decisions and the speed at which lagging markets can begin the investment cycle. But beginning is not completing, and completing grid infrastructure in a constrained urban environment is a task measured in years. The markets that are pulling ahead in 2026 are doing so because of investments made in 2019 and 2020. The markets that begin serious grid investment programmes in 2026 and 2027, enabled and partly funded by the Tech Sovereignty Package, will be in a position to compete for the investment cycle of 2032 and beyond.
The broader lesson written into European data centre geography by the grid queue crisis is one that infrastructure planners in every sector should carry forward: the development decisions that determine competitiveness in a technology cycle are typically made in the cycle before, when demand is not yet visible and urgency is not yet felt. FLAP-D markets built their dominance on connectivity and enterprise proximity investments made in the 2000s and early 2010s. Nordic and Spanish markets are capturing AI infrastructure investment in the 2020s because of energy and grid investments made before AI became a grid planning variable. The countries that are positioned to lead European digital infrastructure in the 2030s are those making serious, funded, and executable grid investment commitments today not those publishing ambition documents while their connection queues lengthen. CADA provides a framework. InvestAI provides capital. The transmission line still has to be built.
Germany and France: The Missing Momentum
Frankfurt’s Constraints and Berlin’s Ambition
Germany’s position in the European data centre race carries a specific tension: it holds the largest number of data centre facilities in Europe, with 529 facilities representing a leading share of the continental total, yet Frankfurt — the country’s primary market — operates under grid pressures similar to those constraining London and Dublin. The city’s concentration of financial services connectivity, internet exchange infrastructure at DE-CIX, and established enterprise relationships creates a base demand floor that does not disappear regardless of grid constraint. Germany’s National Data Centre Strategy, launched in March 2026, targets doubling national capacity by 2030 with a focus on AI-ready infrastructure — an ambition that explicitly acknowledges the current gap between demand and accessible capacity. The strategy’s requirement for heat reuse, exemplified by Maincubes’ FRA04 facility redirecting thermal output into municipal heating grids, reflects a policy approach that attempts to resolve the grid demand tension by turning data centre energy consumption into a network benefit rather than a pure draw. Behind-the-meter power models, such as the Argaman Group’s Frank Cube project in Birstein powered by a dedicated wind farm, remove dependency on the public grid entirely — a model that works for greenfield campuses in locations with developable land and renewable resource proximity, but that cannot be retrofitted onto the constrained urban environments where most of Frankfurt’s existing capacity sits.
France’s Paris market faces a comparable set of constraints in a different political context. The French government’s classification of data centres as strategically important infrastructure creates planning priority without creating physical grid capacity. Paris operates at high capacity utilisation, and the same connectivity-driven base demand that anchors Frankfurt prevents meaningful capacity migration out of the city. France’s nuclear power base gives it a grid carbon intensity advantage that distinguishes it from coal-heavy European grids, but that advantage does not translate directly into connection availability for new large-scale demand. The IEA’s November 2025 analysis of European data centre energy constraints noted that Dublin and Amsterdam have paused new projects, while Copenhagen and Milan have gained significance — and France, despite its nuclear grid advantage, does not feature in the accelerating column. Secondary French cities, including Lyon and Marseille, hold more development potential than Paris itself, driven by the same logic that is redirecting demand from Amsterdam to Eemshaven and from Dublin to unconstrained Irish grid regions.
The Long View: Civil Engineering as Competitive Policy
The grid queue crisis in European data centres carries a lesson that extends well beyond digital infrastructure. Physical systems — electricity grids, water networks, transport corridors — determine the geographic distribution of economic activity in ways that policy frameworks can incentivise but cannot shortcut. The countries that built transmission capacity, upgraded substations, and invested in grid flexibility before AI-driven demand arrived are now receiving capital that their neighbours are losing. The European Commission’s June 2026 Tech Sovereignty Package acknowledges this reality explicitly in its Strategic Roadmap for Digitalisation and AI in Energy, which links data centre development directly to grid investment planning and sets timelines for energy performance standards that will shape the development economics of the sector through the rest of the decade. CADA’s data centre acceleration zones can reduce permitting friction where grid capacity exists. The Roadmap can align energy planning with digital infrastructure demand. The InvestAI funding envelope can provide capital for the grid investments that underpin connection certainty. What none of these instruments can do is substitute for physical infrastructure that simply does not exist yet in markets where years of underinvestment have created the queues that now determine which countries win.
The data centre race in Europe is not primarily a race for talent, tax incentives, or land. It is a race for grid connections, and the podium positions were largely determined by infrastructure investment decisions made before the race began. That is not a counsel of despair for lagging markets — it is a precise diagnosis of what those markets need to do, and how long it will take. Grid investment programmes funded in 2026 and 2027 under frameworks that CADA enables will begin to yield connection capacity in the early 2030s. Markets that make those investments now will be competitive for the next wave of AI infrastructure demand. Markets that wait for the next demand crisis to force the decision will find, as Dublin and Amsterdam found, that by the time the urgency arrives, the capital has already moved.
