Data Desk

Steel's Long Descent Stops on a High Plateau

Steel spent years sliding from the highest prices in the history of the series, and buyers learned to build the slide into their quotes. The slide has stopped, and where it stopped should change how steel-heavy work is priced.

The Bureau of Labor Statistics index of what domestic mills charge for steel reads 374.20index (1982=100) as of Jul 2026, up 22.5% from a year ago. Measured against December 2021, when the series printed 449.711, the highest reading in its archive, today's level is 16.8% lower. That descent is the steel story most buyers have internalized: prices broke, pressure eased, attention moved on to other line items. The archive tells a second story that far fewer people have priced. Even after the full slide, the index sits at the 79th percentile of everything this series has recorded since 1990. Steel did not get cheap. It stopped getting less expensive, and it stopped at a level the archive calls high.

The descent everyone priced in

The series behind that claim is the producer price index for steel mill products: the price at the mill gate for the sheet, plate, bar, and structural shapes that go on to become frames, chassis, enclosures, and building steel. It is a transaction price where the metal is made, not a service-center quote, so it tends to move before the number on your purchase order does. It is also one of the longest-running cost benchmarks a manufacturer can consult, carrying 37 years of monthly history, enough to contain every steel market a working plant manager has ever bought through. That depth is what lets the archive say something the last invoice cannot.

Since year-end 2021 the net move is a decline of 16.8%, and a decline that persists for years does something subtle to the people who buy against it: it hardens into an assumption. Estimators shade steel a little lower in each new quote. Escalation clauses sit unexercised because invoking them feels like arguing against gravity. The material drifts down the list of things worth a meeting. What the assumption misses is the base the fall started from. December 2021's 449.711 was not a normal month to descend from; it was the highest reading the series had ever produced. A retreat from the top of a 37-year archive can run a long way and still leave prices historically elevated, and that is exactly what this one did.

Steel mill products PPI, Jul 2026: 374.20index (1982=100). The archived history runs from 291.56 in November 2025 to 374.20 in July 2026. Today's reading sits at the 79th percentile of the monthly readings since 1990.

Locating the plateau in 37 years of data

Percentile is the honest way to locate a price, because it makes the archive do the work instead of the adjective. At the 79th percentile since 1990, the current reading is not an emergency, but it is nowhere near normal either: most of the months this series has ever printed came in below today's level. The floor of that history is 98.3, set in January 2002. The current index runs 3.8 times that floor. Nobody is forecasting a return to those prices, and the comparison is not offered as nostalgia. It is offered because the mental model many buyers still carry, that steel is working its way back toward some older normal, finds no support in this series. The index stopped falling far above the range where it spent most of its history. A plateau at the 79th percentile is not a correction completing. It is a new price level settling in.

The descent from the peak gave back a slice of the runup. The plateau quietly kept the rest.

What the plateau does to a steel buy

Put dollars on it. Take a fabricator spending $500,000 a year on steel mill products at current prices, and let the index rescale that spend across time. At the December 2021 peak, the same tonnage would have cost about $600,892, so the descent has handed this buyer roughly $100,892 a year of relief against the worst month in the archive. That is real money, and it is the reason steel has slid down the agenda in cost reviews. Now run the same arithmetic against the floor. At January 2002 prices, the identical tonnage would have priced out near $131,346. The gap between that figure and today's $500,000 spend comes to $368,654, which is 3.7 times the size of the relief the descent delivered. Both numbers are true at once. Which one governs your pricing depends entirely on which reference point you choose, and choosing the peak because it flatters the trend is how margin leaks out of steel-heavy work.

One caution this desk owes you: the index records what mills charged, not why they charged it. Tariff regimes, energy costs, mill consolidation, and demand cycles all get credit for the plateau in trade-press explanations, and the series itself confirms none of those stories. This page will not dress a theory in numbers the archive does not contain. What the data does confirm is narrower and more useful: the level. Whatever its cause, 374.20index (1982=100) is the price environment every steel-heavy quote inherits today, and the burden of proof now rests with anyone whose pricing quietly assumes the descent will resume.

What to do with the number

Start with the contracts. Any agreement with a steel escalator anchored near year-end 2021 has been drifting in one party's favor ever since, and if the drift runs against you, the plateau is the argument for reopening it: the reference price no longer describes the market. Next, reprice the work. Run your actual steel spend through the material price variance calculator against the live index rather than last year's invoices, and let the variance, not habit, decide which jobs get re-quoted. Then revisit the boundary decisions. A make-versus-buy analysis done near the peak overstated the cost of purchased steel parts, and one done in the middle of the descent understated it. A plateau is the first stable base that math has had in years, which makes this the moment to redo it. The chart above updates with every BLS release. The decisions priced off it should update just as often.

Feed your actual steel spend and the live index into the material price variance calculator to see what the plateau adds to every job you quote. Reprice your steel line

Published 2026-08-18.