The projection reframes how investors should view energy infrastructure. Utilities, natural gas generators, and nuclear operators are no longer just defensive holdings. They are direct beneficiaries of AI buildout capital.
Three investment themes are accelerating in response.
Power generation capacity: Capital is flowing into generation assets able to supply firm, around-the-clock power. Data centers cannot operate on intermittent renewables alone. Natural gas peakers and nuclear baseload facilities are attracting M&A interest from both utilities and data center developers seeking power certainty.1
AI infrastructure REITs: Specialized REITs holding data center land, facilities, and power interconnection rights are repricing to reflect scarcity in constrained geographies. Nevada, Virginia, and Texas — states where grid saturation is already measurable — are seeing development slowdowns that lift the value of existing, energized capacity.1
Utility margin expansion: Regulated utilities in high-demand corridors are filing for rate increases tied to grid upgrade costs. Investors in these utilities gain exposure to AI-driven load growth without taking direct technology risk.
The margin picture for hyperscalers cuts the other way. Rising electricity costs compress operating margins for the largest data center operators. That pressure creates a bifurcation: companies with long-term power purchase agreements at locked-in rates outperform those exposed to spot market pricing.
Geography is becoming a core variable in data center underwriting. States where grid capacity is exhausted will see development migrate to the Mountain West, the Midwest, and the Southeast — regions with available transmission capacity and lower land costs. Investors tracking utility load growth forecasts in those corridors have an early signal on where the next data center clusters will emerge.
The 2026 Nevada study is an early indicator of a dynamic likely to appear in a dozen states within five years. Power availability — not permitting or labor — is becoming the binding constraint on AI infrastructure expansion.1
Sources:
1 AI Data Center Energy Demand Inflection Study, May 2026
