The grid's physical realities are setting in for the data center pipeline
Latitude MediaLisa Martine Jenkins28 Apr 2026all-rights-reservedRead the original
Site footer carries a 2026 Latitude Media copyright line. Commercial trade journalism, no reuse licence offered. Short quotation with attribution and a link only.
Latitude Media examines how much of the announced US data centre pipeline can physically be connected to the grid. It sets roughly 780 GW of announced data centre projects against a current US peak load of about 759 GW, illustrating that announcements far exceed what the system could plausibly absorb. Against that, Grid Strategies projects around 166 GW of peak demand growth over five years, of which roughly 90 GW is attributed to data centres, while BloombergNEF forecasts about 106 GW of data centre demand by 2035. The article argues that transmission build times, not capital or chip supply, are becoming the binding constraint. It cites Aurora Energy Research modelling suggesting that adding 10 GW in ERCOT without accompanying transmission would quintuple congestion costs and raise prices about 34 percent, and reports contractors extending groundbreaking timelines from six months to eighteen on labour constraints.
Quotation
“"If you want to build the transmission system that you need for 2030 you really needed to have started yesterday." — Oliver Kerr, Aurora Energy Research, quoted in Latitude Media”
Latitude Media · all-rights-reserved · Read the original
All rights reserved. Only this short quotation is reproduced here.
The MyGreenSuit briefing
Commentary by MyGreenSuit, not the words of Latitude Media.
Announced data centre projects now total more than the entire US peak load. That comparison does a lot of work here, and it is the right one — it makes the gap between announcement and delivery impossible to hand-wave.
The number that should change behaviour is Aurora's. Adding 10 gigawatts in ERCOT without matching transmission quintuples congestion cost and raises prices around 34 percent. Congestion is the mechanism by which a speculative interconnection request becomes a line item on somebody's household bill, and the transmission that would prevent it takes longer to build than the data centre does.
The labour detail is quietly the most concrete thing in the piece. Time to groundbreaking has gone from six months to about eighteen, on electrician and contractor capacity. Capital does not fix that quickly, and it is in nobody's forecast.
What this leaves is a community question rather than a technical one: who is at the table when a utility commits to transmission for a pipeline that turns out to be three-quarters vapour, and who pays if it does. So far the people paying have not been the people deciding.