Industries / Energy, water and waste / Electric transmission and distribution / Transformers and switchgear
Cores and windings
The core is the stack of thin steel sheets that carries the magnetic field inside a transformer, and the windings are the coils of copper or aluminum wound around it; how both are made decides how much power a transformer wastes as heat.
6
Milestones achieved, from 1902 to 2024.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1902
Silicon steel cuts losses
An iron core loses power each time the magnetic field reverses, partly through eddy currents, small electric currents that circle inside the metal. Robert Hadfield's iron with about 3.4% silicon, sold as Stalloy, has about four times the electrical resistance of the best transformer iron, which weakens those currents; W. F. Barrett, W. Brown and Hadfield reported on such alloys to the Institution of Electrical Engineers in 1902, as a 1911 Melbourne study that measured Stalloy records. London's Science Museum holds a transformer given by Hadfield that it lists as "the first made of silicon steel".
Credit
Robert Hadfield; W. F. Barrett and W. Brown
1934
Grain-oriented steel
Steel is made of tiny crystals called grains. Norman P. Goss found that cold rolling in stages, with a heat treatment between them, gives a sheet that magnetizes far more easily along the rolling direction; his patent was filed on 7 August 1933 and granted on 3 July 1934. In his own test, a small transformer with the new steel drew an exciting current (the current needed only to magnetize the core) of 2.34 A, against 4.05 A for one built with the best commercial electrical steel, and lost 49.6 W against 51.2 W.
Credit
Norman P. Goss; Cold Metal Process Co. (the patent's assignee)
1972
Steel with sharper grain
Nippon Steel's Orientcore Hi-B lines its grains up more closely than earlier grain-oriented steel, and its engineers first described it in IEEE Transactions on Magnetics in September 1972. In 1974 S. Taguchi, Takumi Yamamoto and A. Sakakura reported core loss two to three grades better than conventional steel and less magnetostriction, the slight stretching of steel when magnetized that makes transformers hum, and said transformers built from it had lower loss, magnetizing power and noise. Nippon Steel now also marks the surface of Hi-B with a laser to split its magnetic domains, the small regions magnetized in one direction, which lowers loss further.
Credit
S. Taguchi, Takumi Yamamoto, A. Sakakura and Takashi Nozawa (Nippon Steel)
1982
Glassy metal cores
Amorphous metal is an alloy cooled so fast, about one million degrees Celsius per second by Metglas's account, that its atoms set in a random pattern instead of crystals. On 13 April 1982 Duke Power switched on a General Electric pad-mounted transformer with a core of Allied-Signal's iron-boron-silicon amorphous alloy in Hickory, North Carolina, and that month Georgia Power installed a Westinghouse unit in Athens, Georgia; an MRS Bulletin history says the alloy cut core losses by 75%. EPRI counts more than 500,000 such units installed in the US over the next 15 years, then says demand there "disappeared" in the late 1990s as deregulation set in and US makers stopped producing them, while India, China and Japan kept buying.
Lessona core that costs more but wastes less sells only where buyers put a price on losses over the transformer's life.
Credit
EPRI; General Electric; Allied-Signal (now Metglas); Duke Power
1997
Windings with no resistance
A superconductor, kept very cold, carries current with almost no electrical resistance, so windings made from it waste far less power as heat. ABB reported in 1997 that a 630 kVA three-phase distribution transformer, rated 18.7 kV to 0.42 kV, with windings of high-temperature superconductor tape from American Superconductor, had been connected to the power network of Geneva, and called it the world's first three-phase transformer of its kind. A 2009 paper by Yinshun Wang and colleagues, who built a similar unit in China, reports that it ran in Geneva's live grid for a year.
Credit
ABB Corporate Research (Baden) and ABB Sécheron (Geneva); American Superconductor
2024
Efficiency rule spares steel cores
By law the US Department of Energy (DOE) sets minimum efficiencies for distribution transformers, the units on poles and pads that serve homes and businesses. DOE says its 2023 proposal would likely have moved about 95% of the market to amorphous cores, while its final rule of April 2024 can be met with grain-oriented steel for about 75% of the market, gives makers five years to comply, and will save $824 million a year in electricity costs. In June 2026, after a presidential finding that transformers and core steel are essential to national defense, DOE asked for comments on how the rule, due in 2029, affects domestic manufacturing.
Lessonan efficiency rule that leans on one core material has to reckon with whether enough of it can be made.
Credit
US Department of Energy
Watch for
Transformers built from chips
A solid-state transformer replaces the iron core and copper coils with power electronics, switching chips of silicon carbide or gallium nitride, so it can also steer power and convert between alternating and direct current. Amperesand, spun out of Nanyang Technological University, said in November 2025 that it would ship 30 MW of medium-voltage units in 2026, starting at the Port of Singapore, and Heron Power and DG Matrix raised $140 million and $60 million in February 2026; TechCrunch reports that the devices still cost more than iron-core transformers. The silicon carbide chips they depend on are the subject of the Note "More chips from a bigger disc".
Achieved whena unit has run for a year in commercial service and its owner or an outside lab publishes its measured efficiency and uptime.
Claimed by
Amperesand, November 2025; Heron Power and DG Matrix, February 2026
Open
01
An open core-loss atlas
A public set of measured losses for grain-oriented and amorphous core steels at 50 and 60 Hz, including the distorted waveforms that inverters and data-center loads put on the grid, measured along and across the rolling direction, with a model trained on it that predicts the loss for a new waveform.
The briefCredit · unclaimed02
A public loss-price table
A table of the prices buyers put on transformer losses when they compare bids: the A factor, in dollars per watt of no-load (core) loss, and the B factor, per watt of load (winding) loss, collected utility by utility from public tenders and filings, each linked to its document and dated.
The briefCredit · unclaimed
Open 01 · the brief
An open core-loss atlas
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of measured losses for grain-oriented and amorphous core steels at 50 and 60 Hz, including the distorted waveforms that inverters and data-center loads put on the grid, measured along and across the rolling direction, with a model trained on it that predicts the loss for a new waveform.
Why it is still open
The open MagNet database, built at Princeton, holds over 150,000 measured waveforms, but for six ferrites at 50 to 500 kHz, the materials of small power-electronics parts. A 2026 Chalmers University paper says makers give data for grain-oriented steel mainly along the rolling direction and at a few frequencies such as 50, 100 and 200 Hz, and that complete measurements in other directions are often unavailable.
Built with what exists
An Epstein frame or single-sheet tester, the standard benches for measuring loss in strips of electrical steel; a programmable power amplifier to make distorted waveforms; a data-acquisition card; MagNet's data format and published method as a template; a small neural network for the model.
First step
Buy Epstein strips of two grades and reproduce each maker's published loss at 1.7 T and 50 Hz.
Done looks like
At least eight grades, including conventional, Hi-B, laser-scribed and amorphous, each measured at 5,000 or more waveforms and published under an open license, with the model predicting held-out losses within 5% and a second lab repeating 50 of the points within 3%.
Open 02 · the brief
A public loss-price table
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A table of the prices buyers put on transformer losses when they compare bids: the A factor, in dollars per watt of no-load (core) loss, and the B factor, per watt of load (winding) loss, collected utility by utility from public tenders and filings, each linked to its document and dated.
Why it is still open
EPRI found in 2009 that A values were "very dispersed nationally," from $2 to $13 per watt, with the low values tied to buyers who had not updated them in years, and it published that range, not a list. The US ENERGY STAR guide to buying distribution transformers says about 22% of US utilities have an A factor of $5 per watt or more, but EPA worked that share out from average retail electricity prices, not from the factors utilities use. Trade articles still work through single assumed values.
Built with what exists
Public procurement portals; state utility-commission dockets; tender documents from cooperatives and municipal utilities; a language model that finds the A and B values in each document and cites the page (see the Note "Findings that cite notes"); a spreadsheet.
First step
Collect 20 public distribution-transformer tenders from US cooperatives and municipal utilities and record their A and B values.
Done looks like
Dated A and B values for at least 100 utilities in at least three countries, each linked, with a second checker agreeing on 95% of entries, and each utility's ratio of A to B, which EPRI says decides how far amorphous cores pay.