The forecast is enormous and the pipeline behind it is thin. MSI Economics, the research unit of construction services and technology firm MOCA Systems, released “Recharging America” on 15 September, projecting U.S. power-generation construction spending roughly doubling to $150 billion a year by 2030 and close to $691 billion cumulatively. The number that matters more sits a few paragraphs in. Against 12 to 16 GW of data-center-related capacity projected for 2026 delivery, MSI estimates about 5 GW is actually under construction.
What the screen actually tested
MSI ran near-term 2026 and 2027 projects against four conditions: a cleared or advanced interconnection queue position, an executed interconnection agreement, reserved transformers, turbines and switchgear, and committed financing. Miss any one, and the project gets pushed to a 2028-or-later start. Equipment reservations were the condition projects failed most often.
That’s the useful part of the exercise, and it’s a discipline most announcement-driven forecasts skip. The $1.66 trillion of data center and advanced-manufacturing investment feeding the model, including $1.25 trillion of data centers, is announced capital, self-reported by the companies announcing it. MSI’s own thesis is that announced doesn’t mean construction-ready.
Lead times are the constraint, not appetite
Large power transformers are quoted at up to 128 weeks, against a pre-2020 norm of 24 to 30 months. Generator step-up transformers run up to 144 weeks. Heavy-frame gas turbines are quoted at five to seven years from order to commercial operation. Interconnection medians for projects that reached service between 2022 and 2025 were 46 months in Texas and nearly 100 months in California. Those two figures are the most count-like numbers in the report.
Supply is concentrated in the same way demand is. MSI estimates imports cover roughly 80 percent of U.S. grain-oriented electrical steel demand, with a single domestic producer on federal records. About 87 percent of the implied construction demand through 2030 lands in ERCOT, PJM and MISO, with Texas alone at $76.2 billion and peak demand projected up 46.8 percent. Solar is about 74 percent of projected capacity additions; gas construction runs near $29.4 billion a year by 2030.
Read the numbers as marketing until someone checks them
MOCA sells construction services and technology, and this is its report. The $150 billion, the $691 billion, the 35 GW and $61 billion annual gap between MSI’s EIA-based floor and its pipeline-based ceiling are all model output. The 5 GW under construction is an estimate, not a count. AGC chief economist Ken Simonson told ENR he cannot independently assess the spending forecast, and nobody in the piece validates any MSI number. MOCA’s own chief economist concedes the upper scenario assumes a takeoff in small modular reactors that isn’t in other baselines.
The behavior it describes is real enough, though. Oracle’s Project Jupiter in southern New Mexico is building a 2 GW-plus on-site power resource, having swapped a gas-turbine-and-diesel scheme for a Bloom Energy fuel-cell microgrid, and issued an RFP on 8 September for 2 GW of renewables for 2027 to 2031 delivery. Broker Howden published its own projection the same day: about 10 GW behind the meter, roughly a quarter of U.S. hyperscale data center power, by 2030. When owners start generating their own electrons, the interconnection queue has stopped being a scheduling item and become a siting decision. Transmission work like the SWIP-North 500-kV line is the other half of the same bet.
ABC’s Anirban Basu offered the one number that settles whether any of this is happening: electrician average hourly earnings are up 7.2 percent over the past year.