Power becomes the schedule

Goldman Sachs Commodities Research forecast in May 2026 that US data-center electricity demand would rise from 31 GW in 2025 to 66 GW in 2027. The more consequential finding was execution risk: only roughly half to three-fifths of scheduled capacity may arrive on time over a one- to two-year horizon.

That turns energization into a first-order investment variable. Site control is not enough. A financeable schedule has to align interconnection, interim generation, natural-gas deliverability, cooling, transformers, turbines, skilled trades, permits, and customer offtake. If one workstream is treated as a late procurement item, the entire revenue calendar can move.

In a September 2026 discussion, Goldman Sachs Research identified turbines, transformers, transmission, tradespeople, policy, price, and physical cooling conditions among the binding constraints on data-center power. The team estimated roughly 30 GW of behind-the-meter gas capacity by 2030, while describing onsite generation as a bridge for projects that ultimately prefer grid connection.

That distinction changes underwriting. A bridge asset must be assessed for its initial duty cycle, fuel and emissions profile, transition path, and residual value after interconnection. A permanent islanded campus and a staged grid-transition project may use similar equipment, but they do not carry the same contract, operating, or capital structure.

Independent operators often need capital to arrive site by site rather than all at once. Crown Point Equity publicly describes HoldCo and per-site SPV structures intended to separate infrastructure financing from the operating sponsor’s balance sheet. Whether any structure achieves its legal, accounting, and credit objectives is transaction-specific, but the organizing principle is durable: capital should follow verified activation milestones rather than assume every megawatt appears simultaneously.