Who makes data center developers, operators & reits for AI data centers

This is the landlord layer: firms that acquire land and power, build or buy the shell and core, and either lease finished capacity to hyperscalers and neoclouds or operate it themselves. It sits upstream of the architecture, contractor, and trade layers below it, which it hires, and downstream of nothing — it's the entity signing the power and land deals that everything else depends on. AI training and inference campuses are denser, larger, and more power-hungry than the enterprise colocation halls this industry was built around, which has pulled a wave of new entrants into the space: bitcoin miners with existing power interconnects repositioning as AI hosts, and private-equity-backed platforms racing to assemble land-and-power pipelines at a scale public REITs can't match on balance sheet alone. For investors, the questions are about the underlying asset, not the building: how much of a developer's pipeline has signed, creditworthy tenants versus speculative land banking; how exposed a given operator is to a single hyperscaler tenant; and how much of the value accrues to the landlord versus getting passed through in the lease. Power interconnection queues, not construction capacity, are increasingly the gating constraint on how fast this layer can add supply.