US steel output spikes to 88.9% utilisation — but the weekly number is an estimate, and the swing is a warning to read it that way

AISI's weekly release for the week ending 31 October 2026 landed at 2.053 million net tons and 88.9% capability utilisation — up 15.7% week on week and 19.2% year on year. Headlines have focused on the year-on-year jump. The more interesting number is the week-on-week one, and the more useful question is whether it describes steelmaking or statistics.

The swing is too large to be routine

The published sequence for the past three weeks is instructive:

Source: AISI weekly raw steel production, as reported by SteelOrbis

A 279,000-tonne rebound in seven days is a 12.1-percentage-point move in utilisation. Blast furnaces do not restart and ramp to full rate inside a week, and even integrated mills coming back from a maintenance outage typically return over a longer window and at partial rate. An EAF-led system is more flexible, but a nationwide 12-point step in a single week — taking utilisation above the mid-October level as well as off the soft 24 October base — is outside the normal range of physical variation.

That points to the structure of the AISI series itself.

What the number actually is

AISI's weekly raw steel report is not a census. The institute describes the tonnage as estimated: weekly production is collected from roughly 50% of domestic capacity, with the most recent monthly data used to cover the remainder. The more complete picture is the monthly "AIS 7" report, based on companies representing 75% of capacity, available by subscription. AISI itself says the weekly release "should be used primarily to assess production trends."

The estimation design has a mechanical consequence: when the reporting panel's data and the stale monthly carry-forward for the non-reporting half re-align, the total can move sharply without any furnace having changed its output. The annualised capacity base implied by the three weeks above is consistent — roughly 2.31 million nt/week, or about 120 million nt/year — which confirms the denominator is stable and the volatility sits in the numerator.

Two further arithmetic points reinforce the case for caution:

• YTD reconciliation. Adding the week (2.053 mnt) to the prior YTD figure (77.405 mnt) gives 79.458 mnt, against AISI's published 79.758 mnt — a 300,000-tonne gap accounted for by AISI's stated adjustment of YTD data to monthly production. The weekly series and the YTD series are not simply additive.
• Run-rate gap. YTD output of 79.758 mnt over 44 weeks implies an average of about 1.813 mnt/week — roughly 78.5% of the ~2.31 mnt/week capability base. The 31 October week ran more than 13% above the year's average. That is either a genuine step-change in the market or a reporting artefact; the balance of evidence favours a large component of the latter.

The underlying trend is less dramatic and more reliable

Strip out the weekly noise and the YTD picture is coherent. Adjusted production through 31 October was 79.758 mnt at 79.5% utilisation, up 6.0% from 75.248 mnt at 77.0% in the same 2025 period. The +6.0% is corroborated by the weekly series' own 44-week averages (1.710 mnt/week in 2025 vs 1.813 mnt/week in 2026), and is consistent with the earlier YTD prints reported through the year — +5.5% as of 19 September, +5.3% as of 3 October. The trend is a modest, steady recovery, not a boom, and the fact that YTD utilisation of 79.5% sits below 2025's 77.0% only because the capability base was smaller is worth watching: output growth is running ahead of the demand signal implied by pricing.

The regional picture matters more than the total

The district split for the week is unusually lopsided and is worth recording because it is less likely to be a pure estimation artefact than the national total (the panel reporting bias should not systematically favour one region):

The Southern district alone produced almost as much as the other four combined (1,001,000 nt vs 1,052,000 nt). That is consistent with the structure of the US industry — EAF capacity concentrated in the South, with Nucor, Steel Dynamics and Commercial Metals accounting for a large share — and with those operators having the flexibility to lift output quickly where order books support it. If any part of the 31 October spike is physically real, it is most likely here.

What would confirm or refute the reading

The next few weekly prints will settle the question. Three things to watch:

1. Whether utilisation holds near 85–90% or reverts to the high 70s. A single-week spike that unwinds within two releases is an estimation effect; a sustained new level is a genuine tightening.
2. Whether quarterly mill guidance corroborates it. Steel Dynamics has been reported running near 92% utilisation against roughly 80% for the industry in Q3; Cleveland-Cliffs' Dearborn Works steelmaking remains idled according to its own filings. Neither points to an obvious national restart explaining a 12-point jump.
3. Whether the Q4 maintenance season shows up in the data at all. The market entered Q4 expecting maintenance outages to remove meaningful supply; a mid-quarter utilisation peak above 88% sits awkwardly against that expectation and is better explained as a statistical event than as deferred maintenance being skipped.

The bottom line for buyers and sellers

The year-to-date trend is the signal: US raw steel output is running about 6% above 2025, at roughly 79.5% utilisation, in a market where flat-rolled prices have been firming on tight supply and higher scrap. The 88.9% weekly print should not be used as a basis for tightening assumptions on its own. AISI's own guidance — assess the trend, not the week — is the correct discipline here, and the 300,000-tonne gap between the weekly arithmetic and the YTD total is a reminder that the weekly series is a directional indicator, not a volume estimate for contracts, inventory or pricing decisions.