I have a public dataset of 365,245 active U.S. manufacturers - just names and locations, no industry codes, no employee counts, no founding dates. The raw counts lie in three specific, fixable ways, and once you correct for them you get a picture of what American manufacturing is: concentrated, rural, and legibly traceable to specific 20th-century optimization patterns rather than the unfiltered list suggests.
These are my findings.
The data is biased toward Illinois, by a factor of two#
Illinois holds 10.5% of the entries in this dataset. According to the Bureau of Labor Statistics, Illinois has roughly 4.4% of U.S. manufacturing employment. Chicago alone has 6,157 entries - more than Houston, New York, and Phoenix combined. If you were going to use this list for market sizing without correcting for that, you'd conclude that Illinois has more manufacturers per capita than any other state by a wide margin, which it does not.
Naming that bias matters because the rest of the analysis depends on knowing it's there. Per-state findings need to be read relative to each state's own internal pattern, not against Illinois.
Per-capita inverts the picture#
Once you normalize by population, the ranking of states with the most manufacturers per 100,000 residents (excluding the over-counted Illinois) looks like this:
- Wisconsin: 184
- New Hampshire: 169
- Iowa: 164
- South Dakota: 162
- Minnesota: 158
- Ohio: 148
- Michigan: 142
- Indiana: 141
- Connecticut: 137
- Oregon: 135
The bottom of the list is the inversion of the top:
- Nevada: 63
- Florida: 67
- California: 68
- Maryland: 71
- Arizona: 73
- Texas: 77
- New York: 83
The four states most commonly cited as American manufacturing powerhouses by absolute count - California, Texas, Florida, New York - are at the bottom of the per-capita ranking. Their absolute numbers are large because the states themselves are large; the share of their economy actually devoted to making physical things is comparatively small.
The real industrial density of America is in the Upper Midwest, the Plains, and the older industrial Northeast. The map that emerges when you correct for population is the Rust Belt plus the dairy belt plus the equipment-and-implement belt, which is to say: it is the map of American manufacturing your grandfather would recognize, not the map that current GDP-share talking points produce.
The "A-1" naming pattern is a fossilized Yellow Pages hack#
215 of the 365,245 companies start with "A-1." Another 185 start with "AAA." This sounded like a Yellow Pages alphabetical-placement hack, so I tested it: do these names cluster in industries where directory placement actually mattered?
The answer is yes, and the effect is large. The categories where A-1/AAA-prefixed names are most over-represented relative to the rest of the dataset:
- Fence and fencing: 12.3x the baseline rate
- Septic, sewer, drain services: 10.9x
- Locksmith and lock/key services: 10.0x
- HVAC, heating, cooling: 4.3x
- Roofing: 4.2x
- Glass and windows: 4.1x
- Plumbing: 3.9x
Compare to industries with no directory dependence:
- Electronics: 0x (zero A-1 electronics manufacturers)
- Construction: 1.0x (baseline)
- Machine shops: 1.4x
- Metal fabrication: 1.5x
The hypothesis lands cleanly. A-1 names cluster in exactly the categories where, in 1985, a homeowner with a flooded basement opened a paper directory and called the first listing. Nobody named an electronics manufacturer A-1, because nobody was finding electronics manufacturers via the Yellow Pages. The naming pattern is industry-specific, not random, and the specificity is the signature of a real optimization target - one that no longer exists, attached to companies that still do.
6.4% of the dataset is national chains, not factories#
Twenty-one entities in this dataset have over 100 separate location entries. The top ones:
- FedEx Office & Print Center: 915 entries across 626 cities
- Fastenal Co.: 711 entries across 678 cities
- Ferguson Enterprises: 573 entries
- Hanger Prosthetics & Orthotics: 367 entries
- Airgas USA: 346 entries
These are distribution centers, retail outlets, and service locations - not manufacturing establishments in any useful sense. If you're using this list to estimate the size of the American industrial base, you need to strip the top 800 multi-location entities, which collectively account for 23,381 entries or 6.4% of the total.
This kind of contamination is invisible in summary statistics but distorts every per-region calculation. Once removed, the per-capita state ranking shifts slightly but the top of the list doesn't change. The Sun Belt giants lose a little more ground because chain locations track population, not industrial activity.
American manufacturing is unusually rural#
This is the finding that surprised me the most, and the one that holds up across every correction. 5,123 cities in the dataset have exactly one listed manufacturer. That's 25% of all cities. These are places like Hazel, South Dakota (Contract Pros Manufacturing); Cleghorn, Iowa (Wetherell Manufacturing Co.); Geyser, Montana (R & L Machine, LLC); Shoshoni, Wyoming (Hoppermetals, Inc.); and Walthill, Nebraska (Tuff-Bilt Tractors Manufacturing, Inc.).
The long-tail concentration goes the other direction from what people usually assume. The top 10 cities - the Chicagos and Houstons - hold only 6% of the dataset. The top 100 cities hold 21.4%. Almost 80% of the entries are spread across 20,250 cities, the median of which contains 5 manufacturers. This is the opposite of the agglomeration-economy picture you'd expect if you'd absorbed the dominant story about manufacturing migrating to a few large clusters. Real American manufacturing is mostly small shops in small towns, distributed widely.
This matters for anyone trying to source domestically, plan reshoring, or build a marketplace on top of this kind of supplier base. The cost of finding a competent shop in your industry is dominated by the cost of locating one of the 20,000 cities where the right shop lives. Every existing platform - Xometry, ThomasNet, MFG - is fighting the long-tail problem first and the demand-matching problem second. The list confirms why.
A Conclusion#
The geographic center of mass of American manufacturing, normalized by population, is the Upper Midwest and Plains. Not California, not Texas. This is not a forecast - it is what the data shows after correcting for an obvious source bias.
Most American manufacturing is rural and diffuse. The cluster story is largely a story about visibility, not about where the shops are.
Co-authored with Claude for data analysis and grammar fixes.