Following the Constraints to the Next Wave

Every advantage that built the established hubs still matters, but one has become the master constraint that now overrides the rest: power. The story of where capacity is heading next is the story of the old clusters running into a power wall and capital flowing to wherever large electricity load can actually be delivered on a workable timeline.

Begin with the wall itself. Northern Virginia, the world's largest market, has grown to the point where its utility faces enormous new load requests and the transmission system cannot absorb them quickly. Interconnection, the process of connecting a large new load to the grid, now runs multiple years in the busiest markets, and in some the queue is so long that new large requests effectively cannot be served soon. Add local pushback over land, water, and the visual footprint of sprawling campuses, and the friction of building in the incumbent hub has risen sharply. The hub does not stop growing, but its growth rate is capped by physics and politics it cannot quickly change.

When the incumbent saturates, capital does not wait; it migrates to where the binding constraint is loosest. The pattern to watch is capacity flowing toward places that can deliver hundreds of megawatts, and increasingly gigawatts, of power soon, even where other attributes are less than ideal. That has lifted a set of emerging US markets, central Ohio, the area around Columbus, which pairs available power with tax incentives and land; the region around Phoenix, Arizona, despite its heat, because power and permitting are workable; parts of Texas, drawn by an independent grid that can move quickly and by abundant gas and wind; and the upper Midwest and Mid-South where utilities have generation headroom. What these share is not perfect climate or the deepest fiber but the ability to actually energize a large load in a reasonable time.

The AI wave has intensified this dynamic because accelerator clusters are far denser and more power-hungry than traditional cloud racks. A single large AI training campus can demand as much electricity as a small city, which pushes operators toward locations near large, dispatchable generation, and toward novel arrangements: co-locating directly at power plants, contracting for dedicated generation, reviving nuclear capacity, and building behind-the-meter gas. When you see an operator siting a campus next to a power station or signing a large direct power-purchase agreement, read it as power scarcity actively reshaping geography, capacity going to the electrons rather than waiting for the electrons to come to it.

Internationally the same logic redraws the map. In Europe, the established cluster of Frankfurt, London, Amsterdam, Paris, and Dublin has hit power and, in Amsterdam and Dublin, explicit constraints on new grid connections, pushing growth toward the Nordics, with abundant hydro power and cool climate, and toward southern and eastern European metros with grid headroom. In Asia, land- and power-constrained Singapore, which paused new capacity for a period, has pushed growth into Malaysia's Johor region just across the border, a textbook case of capacity spilling from a saturated hub to an adjacent one that can supply what the hub cannot. The mechanism is identical everywhere: capacity leaves where the constraint binds and lands where it does not.

So to see the next wave, follow the constraints, not the current rankings. Ask where large power is becoming available, through new generation, transmission upgrades, generation retirements that free headroom, or a grid that can simply move faster, and pair that with permitting speed, tax treatment, and enough fiber to matter. The places where deliverable power is emerging and the other factors are workable are where the next hubs will form, and because power projects and transmission take years, you can often see those conditions assembling well before the capacity actually lands. The constraint is visible in advance, which means the wave is too, if you are reading the geography rather than the headlines.