Digital economy is built on two inseparable forces: capacity and interconnection. On one side, companies, hyperscalers, and cloud providers require megawatts and square meters to host ever-growing loads — AI, cloud native, real-time analytics. On the other side, they depend on dense ecosystems of carriers, IXPs, clouds, submarine cables, and partners moving data at the speed of business. Viewing them as opposing priorities is a common mistake: without capacity, there’s no scale; without interconnection, there’s no performance.
The most iconic carrier hotels worldwide — from 60 Hudson Street (New York) to 350 East Cermak (Chicago), or 1 Wilshire (Los Angeles) — have achieved this status because they do both at once. Today, this balance is reshaping the European map, with Madrid climbing the global data center rankings.
“Operational reality reminds us daily: a data center with only megawatts is an expensive warehouse, and one with only fibers and IXPs is a dead end. The value lies in crossing capacity and interconnection within the same ecosystem,” notes David Carrero, co-founder of Stackscale (part of Aire group), a European provider of private cloud and bare-metal services, working with reference data centers across the continent (Stackscale doesn’t build or operate data centers; it offers services on top of them).
Interconnection: the heartbeat of infrastructure
Interconnection transforms a server building into a living connectivity market. Companies don’t just rent space and power; they seek out the carrier hotels for the ecosystem contained within:
- Low latency: cross-connects and direct peering bypass congested public routes.
- Cost efficiency: peering and fabrics reduce dependency on expensive IP transit.
- Resilience: Multiple carriers and routes — often supported by submarine cables — provide physical redundancy.
- Reach: connect to existing partners in the building to open markets in hours, not months.
In North America, operators like Digital Realty and Equinix have turned their flagship buildings into magnets for carriers, cloud on-ramps, and submarine systems. They extend this value through neutral fabrics at the application level, enabling virtual interconnection across geographies and providers.
Capacity: the foundation for true growth
Interconnection drives performance; capacity sustains growth. A modern carrier hotel must deliver both:
- Power density for AI, HPC, and data-hungry cloud loads (today, 50–80 kW per rack is no longer exceptional).
- Scalability: grow without leaving the ecosystem (same campus, same cross-connect fabric).
- Resilience: A/B power feeds, liquid cooling when needed, maintenance without service interruption.
For years, 350 East Cermak (Chicago) and 56 Marietta (Atlanta) combined capacity and interconnection under one roof. Today, even hyperscalers deploying tens of MW in new developments seek to anchor part of their footprint in interconnection hubs: without an ecosystem, capacity loses strategic value.
Data Gravity: where data attract applications, networks, and partners
The Data Gravity concept changed how we deploy IT. Data — once centralized on on-premises — now grows exponentially in clouds, edge, and hybrid environments. As it grows, it pulls in applications, services, and analytics around it. Practically speaking, the residence of data defines where the ecosystem must live.
Classic examples in the U.S. include:
- 60 Hudson (New York) retains traders and low-latency providers not just for capacity, but for the concentration of brokers, feeds, and carriers at the same point.
- 1 Wilshire (Los Angeles) connects content and CDNs with Asia-Pacific through its submarine window.
- Westin Building Exchange (Seattle) anchors trans-Pacific cables feeding the Pacific Northwest.
The lesson is transferable to Europe.
Europe in perspective: from FLAP-D to Madrid, Marseille, and beyond
The European map was historically dominated by FLAP-D —Frankfurt, LONDON, Amsterdam, Paris, and Dublin. By 2025, this map expands for a pragmatic reason: power availability. With quotas or electricity moratoria in some hubs, capacity and interconnection have grown in cities capable of providing megawatts and fiber backbone simultaneously. Notable examples include:
- Madrid: transitioned from an “alternative” to a main hub: electric capacity with upgrade potential, peering with IXPs (DE-CIX Madrid, ESpanix), backbones, and proximity to submarine cables landing in Bilbao (MAREA, Grace Hopper), Barcelona (2Africa), and Valencia/Cartagena (new systems under deployment). The Spanish capital offers low latency toward the Peninsula, southern France, North Africa, and good transit times to Central European hubs.
- Marsella: Established as the major Mediterranean gateway to Africa, the Middle East, and Asia. Its interconnection ecosystem grew alongside submarine systems converging on the Provençal coast.
- Lisbon/Sines: The EllaLink cable and new projects make Portugal’s Atlantic coast a connection point for Europe–Latin America flows, with competitive land and energy costs.
- Milan/Turin: Northern Italy leverages industry, regional peering, and intra-European connectivity.
- Warsaw: The nexus of Eastern Europe, gaining prominence due to geopolitics and expansion towards the Baltic states.
- Zürich and Vienna/Bratislava: Attractive options for stability, energy, and cross-border links.
“Madrid stands out because it combines megawatts and a rich ecosystem. Adding DE-CIX/ESpanix, proximity to cables, low latency to Iberia, and robust routes to Central Europe creates the perfect environment for high-performance AI and bare-metal deployments,” according to Carrero.
European “carrier hotels”: where peering beats
Europe hosts flagship interconnection buildings that, like their American counterparts, blend ecosystems and scale:
- Telehouse Docklands (London) and Harbour Exchange concentrate peering (LINX) and cloud on-ramps.
- Camps de Interxion/Digital Realty and Equinix in Frankfurt — home to DE-CIX — are central references for Europe’s hub.
- Voltaire/TH2 (Paris) and surrounding campuses aggregate continental transit.
- AMS-IX (Amsterdam) and the Science Park data centers maintain their historic roles.
- MRS1-3 (Marseille): the Mediterranean cable gateway.
- Madrid is expanding across campuses in Alcobendas, San Fernando, Valdebebas, and nearby rings, interconnected through IXPs and backbones.
The pattern is well known: carrier hotel + scale campus within the same metro. The first acts as a market; the second as a megawatt hub. Without bridges (physical and virtual), the equation doesn’t balance.
Submarine cables: the “gravity” crossing oceans
It’s estimated that >95% of international traffic travels via submarine cables. In Europe, this fabric explains why Marsella and the Atlantic coast (from Bilbao to Sines) are growing in interconnection. Deploying in meters with active submarine windows and IXPs reduces latencies, cuts transit costs, and unlocks audience reach in other regions.
Opposition or convergence? The “capacity vs. interconnection” thesis is outdated
By 2025, it’s not about choosing capacity or interconnection, but about convergence. AI, edge, and the global cloud adoption demand infrastructure capable of scaling loads while connecting instantly across the globe. The winning strategy combines:
- Capacity (scalable campuses, power density, liquid cooling) to host pods of 60–80 kW/rack and MW phases of dozens of megawatts.
- Physical interconnection (IXPs, carriers, cloud on-ramps, subsea) and virtual fabrics (neutral fabrics) to move data with predictable latencies and optimized costs.
North America and Europe: two sides of the same coin
The classic article lists North American carrier hotels —60 Hudson, 1 Wilshire, Westin, 350 E. Cermak, 56 Marietta, 11 Great Oaks, 32 Avenue of the Americas— and the operators behind them (Digital Realty, Equinix, CoreSite, Cologix). Europe has equivalent —Docklands, FRA, PAR, AMS, MRS, MAD— and providers with a global footprint reproducing the same model: high-density interconnection in the metro core and scalable capacity in adjacent campuses.
The difference in 2025 is that the demand for AI has raised the bar for capacity, and electrical scarcity at some hubs has shifted Data Gravity toward emerging meters: Madrid, Lisbon/Sines, Warsaw, or Turin/Milan. Interconnection still follows data: where it trains and infers, carriers, clouds, and partners tend to cluster.
How to decide: four practical criteria (and one mistake to avoid)
1) Existing ecosystem. Are there carriers, IXPs, and on-ramps in the building? What cross-connects and meet-me rooms are available? Are there fabrics for virtual interconnection metro-to-metro?
2) Subsea and routes. Access to cables with low latency toward your markets? Are routes diverse and with reliable SLAs?
3) Real scalability. Can the provider double or triple capacity within the same metro without disruptive migrations? Do they have ground and MW pipelines?
4) Density and efficiency. Support for liquid cooling (D2C or immersion), 415 V, 80 kW/rack? Do they measure and publish PUE/TUE and renewable usage?
The mistake: choosing “only capacity” in a connectivity desert or “only interconnection” where power is lacking. Data Gravity punishes both extremes.