Market Data

More Manufacturing Inputs Are at Multi-Year Highs at Once Than a Normal Cycle Can Explain

Sweep the full multi-year archive, not just the last year, and 12 of 15 input-cost series sit in the upper third of their range at once, 9 at outright multi-year highs. Normal cost cycles rotate; this does not. A step-change in the cost base is the honest read, with the cheap exceptions named.

Run a sweep across the manufacturing cost stack, and measure each input against its full multi-year archive rather than the last fourteen months, and one fact jumps out: of 15 key input series, 12 sit in the upper third of their range at the same time, and 9 are at outright multi-year highs. Producer prices for copper have more than tripled since November 1993; machinery, castings, and paper indexes are at the top of their six-year records; wages at $30.35/hour keep setting archive highs. A normal cost cycle rotates, one input peaks while another troughs. When this much of the stack is elevated together, it is less a cycle than a reset of the cost base, and treating it as something to wait out is a strategic error.

Why broad elevation means something different

The whole logic of riding out a cost spike rests on mean reversion: prices that rose on a temporary shock fall back when it passes. That works when a single input spikes on its own supply story. It does not work when 12 of 15 inputs are elevated together over a multi-year window, because that pattern is not a collection of independent spikes waiting to reverse. It is a broad, correlated elevation of the cost base, driven by forces, electrification, energy, structural demand, that are not temporary. The breadth of the elevation, measured over years and not months, is the evidence that this is a level shift, not a blip.

The exceptions, named honestly

Breadth is not unanimity, and pretending it were would be the dishonest version of this argument. Natural gas is the clear exception, still sitting in the lower third of its multi-year range and genuinely cheap by the standards of 2021 and 2022. That is exactly why the multi-year lens matters: over the last fourteen months alone, several of these looked pinned to a high they do not hold over five or six years. The honest claim is not "everything is at a record." It is that a clear majority of the stack has stepped up to the top of its multi-year range while only a few have not, and that majority is too broad to plan around as a passing spike.

The honest scope of the claim

Precision matters, so here is the exact claim: these are highs within the archive we keep, which reaches back to 2020 for the price indexes and 2021 for most other series, not all-time-in-history records, and not a statement about the 2010s we do not hold data for. But the finding holds within that window: a majority of independent input series sitting near the top of their multi-year range at once is a correlated, broad-based elevation, not a set of separate spikes. The strategic implication does not depend on calling it a record. It is not "wait for it to come back down."

One input at a multi-year high is a spike to ride out. Most of the stack at the top of its multi-year range at once is a new floor, and the plants still budgeting for a return to the old one are budgeting for a world that ended.

What planning for a reset looks like

If the cost base has stepped up rather than spiked, the response changes. Reprice quotes to the new level and stop anchoring on stale figures. Build escalation clauses into anything fixed and long. Push hardest on the internal levers, yield, scrap, energy intensity, that lower cost regardless of where inputs sit, because they are the only savings a reset cannot take back. And stop treating each budget cycle as a chance for prices to normalize; plan for the level to hold and be pleasantly surprised if it eases, rather than the reverse. A cost reset punishes optimism and rewards the plants that accept the new number.

Track the input series and the composite pressure index built from them on the live signals page. See the full cost stack

Published 2026-08-06.