The Transformer Queue Starts at One Steel Mill in Pennsylvania

1The queue behind the queue

Grid interconnection is the slowest step in almost every large project in America - the datacentre buildout our bottlenecks board tracks, renewables, reshored plants. Inside that queue sits a harder one: large power transformers, the house-sized units that step voltage between transmission and use. Quoted lead times stretched from under a year a decade ago to roughly three to five years now, and the Department of Energy has repeatedly flagged the fleet's age - much of it past its design life - as a national vulnerability.

Demand has three stacked drivers: new AI load, electrification of everything else, and replacement of a fleet installed in the 1970s. Supply has one deep constraint, and it is metallurgical.

2Grain-oriented electrical steel

A transformer core is stacked from grain-oriented electrical steel - GOES - a silicon steel whose crystal grains are aligned during rolling and annealing so magnetic flux travels with minimal loss. It is one of the hardest commodity steels to make well, and most of the world's capacity sits in Asia and Europe. In North America exactly one producer remains: Cleveland-Cliffs, at the former AK Steel works in Butler, Pennsylvania, with finishing in Zanesville, Ohio.

That single-source status is why GOES shows up in trade actions and why transformer makers hedge with imports. A serious problem at Butler would not stop transformer production worldwide - but it would push the US further into an import queue that is already the bottleneck.

3Who is building capacity

Transformer and electrical-equipment makers have announced US expansions through the decade: GE Vernova (GEV), Hitachi Energy, Siemens Energy, with Eaton (ETN), Hubbell (HUBB) and Powell (POWL) across the broader electrical stack, and Quanta (PWR) building what connects it all. Expansion does not break the constraint quickly: a transformer plant takes years to build and its workforce - coil winders are a craft - takes as long to train.

This is the rare bottleneck where the listed exposure is broad and liquid. The whole layer has repriced since 2022, which is exactly what our bottlenecks board would predict for a constraint this visible; the open question is duration, not existence.

4Watchpoints

Falsifiable things to track: quoted LPT lead times (utility procurement commentary, DOE reports) - the thesis weakens if they compress toward 18 months; GOES trade-case activity and any Butler capacity news from Cliffs; and hyperscaler capex guidance, which is the marginal demand unit. Our macro desk's rate risk matters too - this is long-duration industrial capex, and the brief currently flags the 30-year at 5.12% as a stress level for exactly this kind of spend.

Sources and method

Compiled from public filings, government and agency data and trade press, cross-checked against our desks where they overlap. Supply shares and lead times in private industrial chains are estimates, are labelled as estimates, and the ones we could not verify this week are under what we do not know rather than asserted. The lines below are the factual spine.

• Cleveland-Cliffs' Butler, PA works is the sole North American GOES producer (company and trade-case record)
• LPT lead times are quoted in years, not months, across utility procurement commentary (DOE and industry reporting)
• Macro desk brief 2026-07-28 flags 30Y at 5.12% as a duration stress - the financing backdrop for this capex

Underlying sources

• US Department of Energy - large power transformer and grid reports
• Cleveland-Cliffs disclosures (Butler/Zanesville GOES operations)
• Manufacturer expansion announcements (GE Vernova, Hitachi Energy, ETN, HUBB)
• Parallax bottlenecks board; macro brief 2026-07-28 (30Y at 5.12%)

Written by Parallax Research's research writer on 2026-07-29 and reviewed against the dossier it was given. Figures are as of that date and are not updated in place. Our scoring model's out-of-sample AUC is published in the whitepaper; nothing here is a forecast.