While boards spent this year stress-testing their exposure to AI disruption, infrastructure owners underwent the same diligence and emerged repriced upward.
The numbers are no longer subtle. Power and utilities M&A totaled $216bn across 23 transactions in the six months to May 2026, up 173% from $79bn in the same period last year. NextEra’s $67bn acquisition of Dominion is the largest announced regulated utility deal. BlackRock GIP and EQT took AES private for $49.6bn. Megadeals above $5bn rose from six in 2024 to twenty in 2025.
The driver is straightforward. The five largest hyperscalers have committed between $700bn and $900bn in capex in 2026, and none of it functions without power, grid capacity, and land. The financing market has already repriced accordingly: data center debt issuance nearly doubled last year to $182bn, and hyperscalers issued $121bn in bonds, compared with a five-year average of $28bn.
Infrastructure clears AI diligence for a simple reason, it never competed on the thing AI compresses. A grid, a toll road, or a fiber network earns its return from control of a hard-to-replicate physical asset and long-dated contracted cash flows. There’s no margin for a model to automate away.
But the next phase won’t reward ownership indiscriminately. Interconnection queues in Northern Virginia, Phoenix, and Dallas now run four to seven years. Transformer lead times have stretched to 36–48 months. Gartner expects 40% of AI data centers to be power-constrained by 2027. Premiums will favor owners who can deliver capacity fastest, not merely those who hold it.
Worth saying plainly: this is a leveraged bet on hyperscaler capex holding. That spending is currently outpacing AI revenue, and if it slows, the infrastructure layer reprices, just later and from a higher floor.
The ground everything is built on tends to appreciate quietly. Right now, it’s being priced and financed accordingly.