New Fab Capacity Is Concentrating in 7 Counties Across 6 States — and Local Labor Force Depth Varies Sharply

Summary

New US semiconductor fab capacity is landing in a small, specific set of counties, not spreading broadly: Maricopa County, AZ (TSMC, Amkor), Licking County, OH (Intel), Onondaga County, NY (Micron), Ada County, ID (Micron), Williamson County, TX (Samsung), Utah County, UT (Texas Instruments), and Prince William County, VA (Micron expansion). These seven counties account for essentially all of the CHIPS Act-era megaprojects.

Whether the local labor force is "deep enough" splits sharply by county. Maricopa, Ada, Williamson and Utah counties already carry a real semiconductor-cluster workforce (5,000-30,000 workers in NAICS 334) and above-national bachelor's-degree attainment (34-46%), giving them a technician/engineer base to build on. Licking County, OH and Prince William County, VA are the opposite case: they are adding fab-scale employment (Intel's megasite; Micron's Manassas expansion) onto a county with almost no pre-existing semiconductor employment and, in Licking's case, the lowest bachelor's-attainment rate of the seven (28.2%) - meaning the workforce has to be imported or trained from near-zero, not redeployed from an existing cluster. Nationally, the Semiconductor Industry Association/Oxford Economics project a 67,000-worker technical gap by 2030 (and an independent 2026 estimate puts it as high as 157,000), and TSMC's and Micron's own project delays have been explicitly attributed to skilled-labor shortages - so even the counties with an existing base are not assessed as fully staffed for the incoming scale-up.

Step 1 — Where capacity is actually landing (identifying the counties)

This corpus (askamerica) has no table naming individual semiconductor fab projects or their locations — search_catalog for "semiconductor manufacturing" returned no matching table or column (both terms came back in unmatched_terms). The county list below was built from CRS/industry reporting and cross-checked against BLS county employment data to confirm each county's semiconductor-cluster footprint is real, not just an announcement:

• Maricopa County, AZ — TSMC (three announced fabs, $65B+), Amkor Technology (Peoria, assembly/test)
• Licking County, OH — Intel's New Albany megasite (up to 8 planned fabs)
• Onondaga County, NY — Micron's Clay, NY memory fab campus (up to $6.14B in direct CHIPS funding for this + Idaho)
• Ada County, ID — Micron's Boise Fab 4 expansion
• Williamson County, TX — Samsung's Taylor, TX campus (two fabs + R&D)
• Utah County, UT — Texas Instruments' Lehi, UT wafer fab
• Prince William County, VA — Micron's Manassas fab expansion

These match the CHIPS Act's largest publicly disclosed direct-funding awards (Congressional Research Service R49031 tracks status of CHIPS-funded fabs, though congress.gov blocked a direct fetch this session — see Sources).

Step 2 — Existing semiconductor-cluster depth by county (BLS QCEW, NAICS 334)

Read this carefully: it is current cluster employment, mostly pre-dating the announced megaprojects (2024 QCEW does not yet reflect TSMC/Intel/Samsung/Micron fabs still under construction or ramping). Licking County's NAICS-334 cell is BLS-suppressed both years (disclosure_code='N' — too few establishments to protect confidentiality), consistent with a county that had essentially no semiconductor sector before Intel broke ground; Prince William's tiny existing base (0.21% of private employment) means Micron's Manassas expansion is also landing on thin pre-existing depth, not an established cluster.

Data note: census.cbp_establishments, the more natural table for this cut (establishment counts + employment by NAICS at county grain), was completely unqueryable this session — every query failed with an HTTP 404 on a stale Iceberg snapshot manifest (reported live as a defect). econ.county_qcew was used instead; it carries the same underlying BLS series at finer NAICS detail (down to 334413, semiconductor-and-related-device manufacturing specifically) but several of the smaller counties suppress that 6-digit cell — the 334 (broader electronic-component) level above is the finest level with usable data across all seven counties.

Step 3 — Skill and population depth (ACS 2023 5-year)

Williamson TX, Prince William VA, and Utah UT have the deepest degreed talent pools relative to population (all >42% bachelor's-or-higher, well above the ~35% US average) — consistent with being Austin-metro, DC-metro, and Provo/BYU-adjacent counties respectively. Licking County OH is the clear outlier on the low side (28.2%) and also the smallest population (180,311) of the seven, while hosting what will eventually be Intel's largest US site — the starkest mismatch between announced project scale and pre-existing local skill base in this set.

Step 4 — Is there slack in the local labor market to staff this? (grain-mismatch caveat)

The honest answer requires a grain this corpus cannot supply at the county level: BLS LAUS county-level civilian labor force and unemployment series are not loaded here (econ.regional_employment is state-grain only, confirmed via describe_table/catalog search). At the STATE level, June 2026 unemployment across the six host states was uniformly tight — AZ 4.9%, ID 3.7%, NY 4.6%, OH 3.6%, TX 4.4%, UT 3.6%, VA 3.8% — all at or below the roughly 4.1% national rate. Tight state labor markets mean these counties cannot simply absorb thousands of new fab workers out of existing local unemployment; the slack has to come from in-migration, retraining, or drawing workers out of adjacent occupations, which is a slower and less certain supply than an idle local labor pool.

This state-level tightness is corroborated directly by industry and press reporting: TSMC's Arizona ramp-up and Micron's New York timeline have both been pushed back with skilled-labor shortages cited as a contributing factor, and the Semiconductor Industry Association/Oxford Economics 'Chipping Away' study projects a 67,000-worker national technical gap by 2030 (39% technicians, 41% engineers, 20% computer science) against an industry base of roughly 345,000 workers growing toward 460,000 — with an independent 2026 estimate for the largest-project states running as high as 157,000. Only about 3% of US engineering graduates currently enter the semiconductor industry, which is the structural reason the gap does not close by itself even where local bachelor's-attainment rates (as in Williamson, Utah, or Prince William counties) look strong on paper: a general engineering degree is not the same credential as a fab process-engineer or equipment-technician role, which typically also requires 18-36 months of vendor-specific, on-the-job training.

Bottom line

New capacity is landing in seven counties clustered around six states, and it splits into two groups. Maricopa (AZ), Ada (ID), Williamson (TX), and Utah (UT) counties are building on top of a real existing semiconductor workforce and above-average bachelor's attainment — the labor force is deep in credentials but still not deep enough in the specific technician/process-engineer pipeline, per industry-reported delays and the SIA/Oxford national gap estimate. Licking (OH) and Prince William (VA) counties are the harder case: fab-scale hiring is landing on counties with little to no pre-existing semiconductor employment, and Licking specifically combines the smallest population and lowest bachelor's-attainment rate of the seven — the workforce for Intel's eventual largest US site does not yet exist locally at anything like the needed scale and will have to be built almost entirely from in-migration and new training pipelines.

Every query behind this report

7 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT county_fips, county_name, state_fips FROM geo.counties WHERE (state_fips='04' AND county_name ILIKE 'Maricopa%') OR (state_fips='36' AND county_name ILIKE 'Onondaga%') OR (state_fips='16' AND county_name ILIKE 'Ada%') OR (state_fips='39' AND county_name ILIKE 'Licking%') OR (state_fips='48' AND county_name ILIKE 'Williamson%') OR (state_fips='49' AND county_name ILIKE 'Utah%') OR (state_fips='51' AND county_name ILIKE 'Prince William%')

SELECT area_fips, "year", industry_code, annual_avg_estabs, annual_avg_emplvl, disclosure_code
FROM econ.county_qcew
WHERE area_fips IN ('04013','36067','16001','39089','48491','49049','51153')
  AND industry_code IN ('3344','334413','334')
  AND own_code='5'
  AND "year" IN ('2019','2024')
ORDER BY area_fips, industry_code, "year"

SELECT area_fips, "year", SUM(annual_avg_emplvl) AS total_private_emp
FROM econ.county_qcew
WHERE area_fips IN ('04013','36067','16001','39089','48491','49049','51153')
  AND industry_code='10' AND own_code='0' AND "year"='2024'
GROUP BY area_fips, "year"
ORDER BY area_fips

SELECT county_fips, geo_name, "year", total_population
FROM census.acs_population
WHERE county_fips IN ('04013','36067','16001','39089','48491','49049','51153')
  AND "year"='2023'
ORDER BY county_fips

SELECT county_fips, geo_name, "year", pop_25_plus, bachelors, masters, doctorate
FROM census.acs_education
WHERE county_fips IN ('04013','36067','16001','39089','48491','49049','51153')
  AND "year"='2023'
ORDER BY county_fips

SELECT series, "period", value FROM econ.regional_employment
WHERE series IN ('LASST040000000000003','LASST160000000000003','LASST360000000000003','LASST390000000000003','LASST480000000000003','LASST490000000000003','LASST510000000000003')
ORDER BY series, "period" DESC
FETCH FIRST 70 ROWS ONLY

SELECT series, "year", "period", value FROM econ.regional_employment
WHERE series IN ('LASST040000000000003','LASST160000000000003','LASST360000000000003','LASST390000000000003','LASST480000000000003','LASST490000000000003','LASST510000000000003')
AND "year"='2026' AND "period"='M06'
ORDER BY series