U.S. MANUFACTURING · AUGUST 2026
Where did August’s factory backlog growth come from?
Industries outside transportation supplied about 53.0% of the increase in the measured U.S. factory order backlog in August. They represented only 37.3% of the backlog’s total value. The largest existing backlog and the largest contribution to its latest growth are different questions.
The Census Bureau’s October 2 release reported new factory orders of $663.5 billion, up 0.1% from July, and unfilled orders of about $1.61 trillion. Looking underneath that subdued new-orders headline reveals where outstanding commitments accumulated.
Which industries added to the backlog?
Backlog increase outside transportation
$5.256bn
53.0% of the total monthly increase
Transportation backlog increase
$4.666bn
47.0% of the total monthly increase
Open the complete values and calculation steps
| Coverage | End-Jul 2026 | End-Aug 2026 | Change |
|---|---|---|---|
| Total measured backlog | 1,599.680 | 1,609.602 | +9.922 |
| Excluding transportation | 594.488 | 599.744 | +5.256 |
| Transportation equipment | 1,005.192 | 1,009.858 | +4.666 |
Non-transport share of the monthly increase = (599.744 − 594.488) ÷ (1,609.602 − 1,599.680) × 100 = 53.0%.
Non-transport share of the end-August backlog = 599.744 ÷ 1,609.602 × 100 = 37.3%.
Calculations use Table 3 values published in millions, converted to billions by dividing by 1,000. Shares are rounded to one decimal place. The two component changes sum to the total increase.
Why can backlog rise when new orders barely change?
New orders measure activity during a month. Backlog measures outstanding orders at its end. In Census’s general accounting definition, the ending backlog equals the starting backlog plus new orders net of cancellations, less net sales. A small change in the pace of new orders can therefore coexist with a growing stock of unfinished commitments.
Think of the question in two parts: how large is the queue, and how much was added to it after work moved through? The first bar answers a composition question about the queue. The second answers a contribution question about one month’s net change. Neither bar measures physical factory capacity.
For someone researching an industrial supplier, the next step is to match the supplier’s products to the relevant industry series, then look at shipments and delivery performance. A broad backlog total alone cannot reveal whether an individual business is winning profitable orders or struggling to deliver them.
Does the reported 6.87 ratio mean a delivery wait?
No. Census Table 7 reports a 6.87 unfilled-orders-to-shipments ratio for durable goods, up from 6.81 in July. Its denominator excludes industries without unfilled orders, and the ratio excludes semiconductor manufacturing. Dividing the $1.61 trillion backlog by all factory shipments would use a different universe.
Treat the published ratio as an aggregate value comparison with specified coverage. It is not a promise that an order placed today will arrive in 6.87 months.
Open the main limitations before interpreting a larger backlog
- Prices: these are dollar values, without adjustment for price changes. They do not directly count machines, aircraft or finished units.
- Timing: August is preliminary and July is revised. Seasonal adjustment and late responses can change the estimates.
- Coverage: semiconductor unfilled orders are unavailable. The headline total should not be described as a complete semiconductor order book.
- Uncertainty: Census identifies nonsampling risks, including coverage and response error. Its panel is not a probability sample, so conventional sampling-error estimates cannot be calculated.
Does more backlog signal stronger demand or delivery trouble?
Either may contribute, and the aggregate does not separate them. More new commitments, slower completion, cancellations and changes in contract values can produce different paths. Test an explanation against relevant shipments, capacity evidence and company disclosures before attaching a causal story.
What should be checked in the next release?
Recalculate the contribution shares using the revised prior month, rather than carrying this month’s result forward. A one-month majority is a finding about August, not an established trend or an investment recommendation.
Sources, dates and reuse
Observation: end-July and end-August 2026 backlog, plus August monthly new orders. Release: October 2, 2026, CB26-150. Sources checked October 9, 2026 (UTC). Current-report links can advance to a later vintage; the figures above preserve the August full report.
- Census, October 2, 2026 full M3 release, CB26-150
- Census, August 2026 Table 3: unfilled orders, industry detail and footnotes
- Census, August 2026 Tables 7–8: ratios, exclusions and revisions
- Census M3 definitions: new orders, shipments and backlog
- Census citation and public-use guidance
- Census research transparency policy: copyright qualifications
Original analysis and graphics use selected Census statistical facts. Census is credited as the original data source; calculations and conclusions here are the author’s responsibility. Census guidance says employee-created data and works generally lack U.S. copyright protection, while third-party works and use outside the United States may differ. No Census media assets, logos or third-party illustrations are reproduced.
