Why Capacity Concentrates Where It Does

Data-center capacity is strikingly concentrated. A handful of metros hold a wildly disproportionate share of the world's compute, and the concentration is not an accident of history but the product of a stack of reinforcing advantages. To read where capacity is going, you first have to understand why it piled up where it did.

Northern Virginia is the canonical example, the largest data-center market on earth by a wide margin. Its dominance began with a quirk of internet history: a large share of the early internet's fiber and a major exchange point were built there, so networks converged on the region. Once the fiber and the peering were dense, new facilities wanted to be near them to minimize latency and interconnection cost, which drew more networks, which drew more facilities. That is a classic agglomeration loop, where each addition makes the location more attractive to the next, and it explains why capacity clusters rather than spreading evenly.

But fiber gravity is only the first ingredient. Layer on the others and the recipe for a hub becomes clear. Power is paramount: early hubs sat where electricity was abundant, reasonably priced, and deliverable at scale, and where utilities were willing and able to build the substations and lines a large load needs. Land matters, large contiguous parcels at a workable cost, on stable ground away from severe flood and seismic risk. Climate matters, because a cooler or drier climate lowers the energy and water needed to reject heat, improving operating cost. Fiber and network density, as noted, create the interconnection advantage. And policy, tax treatment of equipment and construction, permitting speed, and local political receptivity, can make or break a location's economics.

Tax and permitting deserve emphasis because they are often decisive at the margin. Several US states, Virginia among them, offer sales-tax exemptions on data-center equipment that materially lower the cost of the single largest capital line, the electrical and mechanical gear. A jurisdiction that exempts that equipment and permits quickly can win projects from a neighbor that is otherwise comparable. Conversely, a place where permitting drags for years or where local opposition blocks projects effectively removes itself from contention no matter how good its power or fiber. Operators route capital to where they can actually build.

Proximity to existing gravity is the final reinforcing factor. Operators cluster near their own and each other's existing facilities because the local ecosystem, the construction crews, the electrical contractors, the trained operations staff, the utility relationships, is already built. Entering a metro where you already operate is cheaper and faster than pioneering a greenfield one. This is why hubs tend to thicken rather than new ones springing up from nothing, and why the biggest markets keep growing even as they strain.

Network topology reinforces the same pull in a way worth spelling out. Traffic between an operator's own facilities, replication between availability zones, backhaul to the global backbone, is cheapest and lowest-latency when the facilities sit near dense fiber and established peering. A campus planted far from that fabric pays more for long-haul transport and suffers higher latency to the rest of the operator's estate. So even when raw land and power look attractive somewhere remote, the network cost of isolation can tip the decision back toward an existing hub or toward a site that at least sits on an established fiber corridor. This is one reason emerging markets that succeed almost always sit on a major long-haul route rather than truly in the middle of nowhere: they need enough network gravity to connect back to the core cheaply, even when their headline draw is cheap power.

The crucial insight is that these advantages are stacked and partly substitutable. A location can overcome a weakness in one dimension with strength in another, cheap abundant power can pull capacity to a place with mediocre climate, and a tax exemption can offset higher land cost. But one factor has become the binding constraint above all others, and it is the one that is now breaking the old clusters and redrawing the map. That factor is power, and the next lesson follows it to where the next wave is forming.