The data center power forecast is growing. A forecast still isn’t a connection date.
New energy projections underline the scale of digital infrastructure demand. Developers still need a site-specific route to power.
Image: AI-generated editorial illustration, not a photograph of a specific project.
INFRADAY ANALYSIS
A national electricity forecast cannot tell a developer when a particular substation will be ready. That gap deserves more attention as data center investment moves from announcements into construction commitments.
The International Energy Agency’s 2026 outlook projects global data center electricity consumption rising from 485 terawatt-hours in 2025 to about 950 terawatt-hours in 2030. Separately, the U.S. Department of Energy’s resource hub cites Berkeley Lab scenarios in which data centers account for 9.5% to 15.3% of U.S. electricity use at the end of the decade. These are projections with different geographic scopes, not measurements of capacity already available to developers.
DOE also includes an important qualification: the equipment-based demand model does not directly resolve how much additional electricity the grid or on-site generation can supply. For a project team, that distinction belongs near the front of the investment memo.
Turn the demand forecast into a delivery sequence
Infraday’s view is that developers should carry at least three separate milestones through underwriting: the first construction supply, the first usable block of operating capacity, and the campus’s eventual full requirement. Treating those as one date makes it difficult to see where the schedule is exposed.
A utility discussion should establish what is committed, which studies remain open, who owns the required upgrades and what happens if the campus grows more slowly than expected. An early indication of available capacity should be recorded as exactly that until its conditions are understood.
Phasing deserves the same discipline. A smaller first phase may reduce the amount of capital waiting for electricity, but only if the first phase can operate commercially on its own. Roads, cooling infrastructure and network connections need to fit that sequence.
Give uncertainty an owner
Investment committees should ask for a downside schedule alongside the preferred one. What spending can be deferred? Which equipment orders become difficult to change? At what point must the customer make a binding commitment?
The larger national forecast strengthens the case for serious infrastructure planning. It does not remove the need to test the particulars of an individual site. A credible project schedule should show how power will arrive, in what increments, and who is accountable for each step.
Sources: IEA, Key Questions on Energy and AI, 2026; U.S. Department of Energy, Data Center Resource Hub.
Based on the public sources linked above, reviewed September 27, 2026. Project schedules and company commitments are attributed to their issuers; analysis and recommendations are Infraday’s.