From announced megawatts to usable power.

The US Department of Energy’s December 2024 summary of Berkeley Lab research estimated US data-centre electricity use at 176 TWh in 2023, with a projected range of 325–580 TWh in 2028. These are national energy estimates, not a schedule of capacity available to individual sites.

First, check the units

Megawatts measure power at a point in time; megawatt-hours measure energy over a period. A hypothetical 100 MW facility drawing its full rated load for all 8,760 hours of a non-leap year would consume 876,000 MWh, or 0.876 TWh. That arithmetic says nothing about whether the connection is ready.

Actual consumption depends on ramp-up and utilisation. The quoted MW figure also needs a boundary: IT equipment load and total facility load are different measures because cooling and other systems consume electricity.

Read a project announcement in sequence

1. Site: identify the location, utility, and whether the capacity figure describes IT load or the whole facility.
2. Connection: look for the service agreement, required network work, and the conditions attached to the delivery date.
3. Equipment: distinguish an order from delivery, installation, and commissioning.
4. Revenue: establish when a customer starts paying and who bears the cost of a delay.

For an investor, the useful bridge is from a project milestone to cash flow. A large demand forecast can support further research, but it cannot establish the return on a particular utility, equipment supplier, or data-centre operator. Contract terms and the cost of completing the project still determine the economics.

Research commentary. Examples explain a method; they are not investment recommendations. Sources were reviewed on 2 October 2026.