Powered land and zombie projects: Real estate in the age of AI

The real tradeable signal here is not “AI demand is strong” but that the market is beginning to price a new bottleneck premium: power access is becoming the scarce asset, not compute hardware or raw land. That shifts value capture away from hyperscalers’ preferred economics and toward owners of pre-powered sites, grid-enabled industrial property, and operators that can stitch together generation, storage, and planning into a financed package. In that regime, every additional month of grid delay increases the embedded option value of already-connected campuses while simultaneously destroying the viability of speculative greenfield projects.

For EQIX, this is a supportive medium-term narrative but not a clean near-term catalyst. The company’s edge is less about AI itself and more about converting balance-sheet strength into de-risked capacity at a time when competitors are stuck in the queue; that should compress competitive intensity in premium metros and raise replacement-cost economics for colocation. The second-order effect is that enterprise customers facing 3-5 year delivery risk may extend lease terms or pay up for interim capacity, which is margin-accretive even before new builds come online.

The bigger, underappreciated implication is negative for land-flippers, pure-play developers without power, and any business model dependent on “planning plus hope.” If queue reform works, the pool of viable projects shrinks sharply, which should widen the gap between powered land and everything else and likely trigger write-downs in speculative portfolios over the next 6-18 months. On the demand side, high UK electricity prices are the key constraint: if power costs stay elevated, hyperscalers will treat Britain as a selective, not wholesale, deployment market, and some announced capacity may be delayed rather than canceled.