US steel output hits 80.9% capability utilisation: reading the 4.1% weekly jump correctly

AISI's week-42 release (week ending 17 October 2026) shows US raw steel production at 1.870 million net tons and capability utilisation at 80.9%, up from 1.796 million tons and 77.8% a week earlier and 1.722 million tons and 75.1% a year ago. Year-to-date adjusted output reached 75.632 million net tons at 79.1%, a 5.3% increase on the 71.805 million tons recorded over the same period of 2025.

The temptation is to read the 4.1% week-on-week gain as a sudden acceleration in mill operating rates. It is worth resisting that. The more durable and more consequential number in this release is the 79.1% year-to-date average against 77.0% last year.

What the weekly number actually is

AISI is explicit about the method: the weekly tonnage is an estimate compiled from production data supplied by companies representing roughly 50% of domestic capacity, combined with the most recent monthly figures for the remainder. AISI itself directs users to treat the series "primarily to assess production trends". The more granular benchmark is the monthly AIS 7 report, built on companies representing about 75% of capacity, and the year-to-date figures are adjusted for consistency with that monthly data.

This matters for interpretation. A 74,000 net ton week-on-week change is entirely plausible as real output movement, but it is equally plausible as sampling and estimation noise within a half-capacity panel. One week of data from this series should not be used to infer a step-change in operating behaviour across the industry.

A second methodological point worth flagging for anyone building models on these numbers: AISI reports a capability utilisation rate, not a nameplate capacity rate. Capability is a survey-based practical maximum and sits below nominal nameplate capacity — which is why backing out an implied capacity base from these figures gives a self-consistent answer that will not match published nameplate totals.

Implied capability base derived by dividing reported tonnage by reported utilisation; AISI rounds utilisation to one decimal, so each implied figure carries roughly ±0.02–0.03 mnt/week of uncertainty. Analyst calculation, not an AISI disclosure.

The signal that survives the noise

Because the weekly estimate is noisy, the year-to-date average is the more reliable read on where the US industry is operating. At 79.1%, it is 2.1 percentage points above the 77.0% recorded over the comparable 2025 period — a meaningful shift in average operating rate, not a one-week artefact. The absolute volume gain is roughly 3.8 million net tons year-to-date.

That lift is consistent with the demand backdrop visible in adjacent AISI and trade-press reporting over September and October 2026: flat and long product prices rising, October scrap settling up $20–30/gt in most districts, Nucor's CSP hot-rolled coil price up for a twelfth consecutive week on thin spot supply, and imports climbing even so. Higher domestic utilisation alongside rising imports is the classic signature of a market where domestic supply is being absorbed faster than it is being added — though the causality cannot be established from production statistics alone.

Geography: the Southern share is the structural story

AISI's district breakdown for the week illustrates the shape of the US industry more clearly than any single utilisation figure:

Shares are the author's calculation from AISI district data.

The Southern district's 45% share reflects the region's concentration of electric-arc furnace capacity running on scrap, direct reduced iron and pig iron. This has a practical consequence for anyone tracking input markets: when the national utilisation rate climbs, the marginal ton is disproportionately an EAF ton, and the first markets to tighten are ferrous scrap, metallics and electrodes rather than iron ore. The scrap price increases reported through October are consistent with that mechanism, though they also reflect other supply factors.

Caveats on units and comparability

All figures are in net tons (2,000 lb). For cross-region comparison, 1.870 million net tons is approximately 1.696 million metric tonnes. The distinction matters when benchmarking against World Steel Association data or non-US producers, which report in metric tonnes; the roughly 10% gap between the two units is larger than most of the year-on-year changes being discussed.

Readers should also note that week 42's 80.9% is the highest rate appearing in the 2026 material reviewed here, but a single weekly peak in an estimated series is weak evidence of a sustained ceiling. The test will be whether the year-to-date average continues to grind higher through the fourth quarter, or whether the weekly rate mean-reverts toward the high-70s.

What to watch

The next informative data point is not the following week's estimate but the monthly AIS 7 release, which covers a substantially larger share of capacity and is the series AISI treats as authoritative. If the monthly data confirms output running near 80% of capability, the year-on-year gains in scrap, metallics and alloy consumption implied by these figures are likely to be real rather than statistical. If the monthly series lands materially below the weekly estimate, the week-42 spike should be treated as noise.