A Kalshi market assigned high odds to more than 5,100 U.S. data centers being planned or operating before 2027. The number sounds precise, but the investment signal is weaker than the probability suggests because “data center” can mean anything from a small enterprise room to a hyperscale campus drawing hundreds of megawatts.
The U.S. Department of Energy offers a more economically useful measure. A December 2024 analysis estimated that data centers consumed about 176 terawatt-hours of electricity in 2023, or 4.4% of U.S. power use, and could reach 325 to 580 terawatt-hours by 2028. That range would represent roughly 6.7% to 12% of national electricity consumption. Even the low end implies a large load increase; the width of the range also shows how uncertain deployment, chip efficiency and utilization remain.
Count versus capacity
Facility counts mix together projects of radically different scale and status. “Planned” may include a public announcement, a land option or a utility interconnection request, none of which guarantees construction. Duplicate proposals and phased campuses can also inflate the count. Conversely, a single hyperscale site may matter more to a utility, turbine maker or fuel-cell supplier than dozens of small buildings.
For investors, megawatts energized, contracted power and capital committed are stronger evidence than a headline count. The timing matters too: a project can secure land years before transformers, generation and transmission are available. EIA expected U.S. electricity sales to rise to about 4,135 billion kilowatt-hours in 2026, roughly 2% above 2025. The data-center forecast therefore points to a growing share of a system whose overall demand is already expanding.
What the market is really pricing
The Kalshi contract is best read as a sentiment gauge for continued AI infrastructure construction. It is not a forecast of revenue for any particular public company. Suppliers capture economics only when projects pass permitting, interconnection, financing and procurement milestones. A proposed campus does not equal an equipment order, and an equipment order does not equal recognized revenue.
The decisive evidence will be utility load forecasts, signed power agreements and actual energized capacity. If those measures keep rising, the investment thesis survives even if the facility count misses 5,100. If projects remain stuck in interconnection queues, the count could clear the threshold while the revenue opportunity arrives much later.