The H-1B $100K fee has barely dented tech hiring/wage aggregates — its footprint is a rounding error against a labor market that was already softening

Summary

Barely, and probably not enough to matter for a sector-wide labor-cost read. The September 19, 2025 proclamation imposing a $100,000 fee on new H-1B petitions filed for workers abroad has clearly crushed the volume of fee-subject H-1B filings themselves — DHS's own figures, surfaced in litigation, show roughly 87% fewer such petitions filed between September 2025 and February 2026 than the year before, with only 85 fee payments collected in that window. But that collapse has not produced a visible break in the broad tech-sector employment or wage series. US Information-sector (NAICS 51) employment fell steadily through all of 2025 and into 2026 — from a ~2.93M monthly average in 2024 to ~2.87M in 2025 to ~2.81M in the first seven months of 2026 — and the rate of decline did not accelerate at the fee's effective date; it was already on that path for a year beforehand. Average pay in Computer Systems Design Services (NAICS 5415, the most H-1B-intensive detailed industry) did jump noticeably in 2025 (+9.3% year over year, versus +8.1% in 2024), but that is consistent with, and mostly attributable to, ordinary labor-market composition effects (AI-driven layoffs concentrated among lower-paid/junior roles) and the accompanying wage-weighted H-1B lottery redesign — not something a warehouse-level aggregate can cleanly attribute to the fee. The mechanical reason the fee cannot move the aggregates much: the entire annual H-1B cap is 85,000 visas, while the Information sector alone hires 65,000-91,000 people in a single month (BLS JOLTS). Even a total wipeout of new H-1B hiring is a rounding error against monthly hiring flow, let alone the ~2.8 million-worker sector stock.

One more fact matters for reading the current data: the fee itself is in legal limbo. A federal judge vacated it as an unlawful tax on June 8, 2026, then stayed that ruling pending the government's appeal (now before the First Circuit), so it remains in effect for now — but DHS has separately proposed a new, very similar $103,265 fee via formal rulemaking (published August 25, 2026, comment period closing September 24, 2026) as a hedge against losing the appeal. Any labor-cost read built on the original proclamation's fee is therefore reading a policy that could be replaced or reversed within months.

1. What changed, and when

On September 19, 2025, a Presidential Proclamation imposed a one-time $100,000 fee on new H-1B petitions filed on or after September 21, 2025, for beneficiaries outside the United States requiring consular processing. It explicitly exempted: petitions filed before the effective date, already-valid H-1B holders, and extensions/amendments/changes of status for workers already in the US (which is how most F-1-to-H-1B transitions work). Penn Wharton Budget Model estimates roughly 60% of the annual registration pool was "fee-subject" under this definition — concentrated among older (avg. age 34.2 vs 29.4), bachelor's-degree-only (11.5% vs 63.8% with a master's or higher), Indian-national (82.2% vs 51.2%) computer-occupation workers with no prior US visa (89%).

The fee's legal status has since become unsettled: a coalition of state attorneys general sued (California et al. v. Trump et al., D. Mass.); on June 8, 2026 Judge Leo Sorokin vacated the fee as an unlawful tax, then days later stayed his own ruling pending the government's appeal to the First Circuit, so USCIS is still collecting it as of this writing. Separately, DHS filed a notice of proposed rulemaking on August 25, 2026 to impose a $103,265 fee on all H-1B cap-subject petitions through ordinary fee-setting authority — a belt-and-suspenders move in case the proclamation-based fee is struck down for good.

2. How much it has shown up in H-1B filing volume itself

At the level the fee directly targets, the effect is large and immediate: DHS figures cited in litigation show the department processed 87% fewer fee-subject H-1B petitions in the Sept 2025-Feb 2026 window than the year-earlier period, and collected fee payments on only 85 qualifying applications — versus an expected pool in the tens of thousands. Employers overwhelmingly chose not to file for candidates who would trigger the fee. Penn Wharton's independent simulation is consistent with this: it estimates only 30-56% of fee-subject registrations would 'survive' the payment (i.e., be worth it to the employer), screening the fee-subject share of the lottery pool down from ~60% to roughly 31-46%.

3. Whether that has shown up in tech-sector hiring (BLS CES/JOLTS)

No visible break at the effective date. Summing BLS CES statewide employment series (econ.state_industry) to a US total for the Information supersector (NAICS 51):

• 2024 monthly average: ~2,930,000
• 2025 monthly average: ~2,868,000 (-2.1% y/y)
• 2026 Jan-Jul average: ~2,811,000 (-2.0% y/y on a like-for-like partial-year basis)

Comparing the eight months before the fee (Jan-Aug 2025, avg ~2,877,000) to the eleven months after it (Sep 2025-Jul 2026, avg ~2,826,000) gives a further -1.8% drop — almost exactly in line with the pre-existing year-over-year trend rather than an acceleration. The Information sector's job-openings and hires levels (BLS JOLTS, econ.jolts_industry) show the same pattern: no discontinuity around September 2025, just ordinary month-to-month churn (hires ranging 65,000-91,000/month across the period; openings 80,000-157,000/month). Professional and Business Services (NAICS 60, which includes IT consulting/staffing where H-1B usage also concentrates) shows the same absence of a visible break — hires ran 900,000-1,090,000/month before and after the fee with no step change.

This is not surprising given the arithmetic: the entire H-1B cap is 85,000 visas per year, and consular-processing (fee-subject) petitions are a subset of that. A month's ordinary hiring in the Information sector alone exceeds the entire annual H-1B cap.

4. Whether it has shown up in wage data

There is a real signal here, but it is ambiguous and cannot be cleanly attributed to the fee from this data alone. BLS QCEW average annual pay in Computer Systems Design Services (NAICS 5415, econ.state_wages, 2.4 million private-sector workers) rose from $160,939 in 2024 to $175,982 in 2025 — a 9.3% jump, compared with 8.1% the year before and single-digit growth in most prior years. Employment in the same industry kept falling (2,438,858 in 2024 to 2,398,982 in 2025, -1.6%). Two mechanisms could both be contributing, and QCEW cannot separate them: (a) ordinary composition effects — 2025's well-documented wave of tech layoffs concentrated at junior/lower-paid levels, mechanically raising the average pay of who remains; and (b) the new wage-weighted H-1B lottery (in effect for the FY2027 cycle) and the fee itself, both of which Penn Wharton's simulation finds shift the selected H-1B population toward higher earners — though PWBM's own estimate is that most of that compositional shift comes from the wage-weighting redesign of the lottery, not the $100,000 fee, and the fee's own marginal effect on selected pay is small and even sign-ambiguous (+6.2% under one assumption about employer payroll savings, -2.9% under another). Because QCEW is an annual average and the fee only covered the final ~3.5 months of 2025, only a fraction of any fee-driven compositional shift could show up in the 2025 figure at all; a full-year 2026 figure (not yet published) would be the first real test.

5. Is it enough to matter for a sector-wide labor-cost read?

No — not at the level a labor-cost analyst would use to model overall tech-sector costs. The fee-subject population is a few thousand to perhaps 20,000-30,000 workers a year at the outside (a fraction of an 85,000 annual cap, further cut ~55-70% by the fee itself per DHS's own filing data and PWBM's survival estimates), against an Information-sector workforce of ~2.8 million and a Computer Systems Design Services workforce of ~2.4 million. Even a full pass-through of the $100,000 fee onto affected hires would add well under 0.1% to total sector payroll costs. Its main real-world effects to date are compositional and micro-level (who gets selected, what nationality/degree mix, some documented offshoring of roles to lower-cost countries per employer reporting — see IEEE Spectrum) rather than macro-level (aggregate headcount or average wage). Two things would need to happen before this changed: (1) the fee (or its $103,265 rulemaking successor) surviving the current litigation and rulemaking process intact, and (2) a full year of QCEW/CES data accumulating under a stable rule, since the current data window captures at most a partial year under a fee whose legal status is still unresolved.

What This Report Does Not Answer

• H-1B-specific microdata (filings, LCA-level detail) not directly queryable in this warehouse: search_catalog found no H-1B/LCA table in this corpus; filing-volume figures (the ~87% decline, 85 fee payments) come from DHS figures reported in litigation and secondary news coverage, not a primary DOL/USCIS dataset fetched directly.

Every query behind this report

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

SELECT industry_code, industry_name FROM econ_reference.jolts_industries WHERE industry_name ILIKE '%information%' OR industry_name ILIKE '%professional%'

SELECT supersector_code, supersector_name FROM econ_reference.naics_sectors WHERE supersector_name ILIKE '%information%' OR supersector_name ILIKE '%professional%'

SELECT industry_code, year, date, metric_type, value
FROM econ.jolts_industry
WHERE industry_code IN ('510000','540099')
  AND metric_type IN ('JOL','JOR','HIL','HIR','QUL','QUR','LDL','LDR')
  AND ((year=2025 AND date >= 'M01') OR year=2026)
ORDER BY industry_code, year, date

SELECT industry_code, "year", "date", metric_type, MAX("value") AS value
FROM econ.jolts_industry
WHERE industry_code IN ('510000','540099')
  AND metric_type IN ('JOL','HIL','QUL','LDL')
  AND (("year"=2024 AND "date">='M06') OR "year"=2025 OR "year"=2026)
GROUP BY industry_code, "year", "date", metric_type
ORDER BY industry_code, metric_type, "year", "date"

SELECT "year", "period", SUM("value") AS total_employment_thousands
FROM econ.state_industry
WHERE series LIKE 'SMU__000005000000001'
  AND "year" IN (2024,2025,2026)
  AND "period" <> 'M13'
GROUP BY "year","period"
ORDER BY "year","period"

SELECT "year","period", COUNT(*) AS n_states
FROM econ.state_industry
WHERE series LIKE 'SMU__000005000000001'
  AND "year"=2026
GROUP BY "year","period"
ORDER BY "period"

SELECT series, "value"
FROM econ.state_industry
WHERE series LIKE 'SMU__000005000000001'
  AND "year"=2026 AND "period"='M05'
ORDER BY "value" DESC
LIMIT 60

SELECT "period","value" FROM econ.state_industry
WHERE series='SMU06000005000000001' AND "year"=2026
ORDER BY "period"

SELECT a.series, a.value AS m04, b.value AS m05, a.value-b.value AS diff
FROM (SELECT series,"value" FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2026 AND "period"='M04') a
JOIN (SELECT series,"value" FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2026 AND "period"='M05') b
ON a.series=b.series
ORDER BY diff DESC
FETCH FIRST 15 ROWS ONLY

SELECT SUM(a.value) AS sum_m04, SUM(b.value) AS sum_m05, COUNT(*) AS n
FROM (SELECT series,"value" FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2026 AND "period"='M04') a
JOIN (SELECT series,"value" FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2026 AND "period"='M05') b
ON a.series=b.series

SELECT "period", COUNT(*) AS n_rows, COUNT(DISTINCT series) AS n_series
FROM econ.state_industry
WHERE series LIKE 'SMU__000005000000001' AND "year"=2026
GROUP BY "period" ORDER BY "period"

SELECT SUM("value") FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2026 AND "period"='M04'

SELECT "year","period",SUM("value") AS tot FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2026 GROUP BY "year","period" ORDER BY "period"

SELECT "year","period",SUM("value") AS tot FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2025 GROUP BY "year","period" ORDER BY "period"

SELECT "year","period",SUM("value") AS tot FROM econ.state_industry WHERE series LIKE 'SMU__000005000000001' AND "year"=2024 GROUP BY "year","period" ORDER BY "period"

SELECT "year","period", AVG("value") AS avg_hourly_earnings
FROM econ.state_industry
WHERE series LIKE 'SMU__000005000000003'
  AND "year" IN (2025)
GROUP BY "year","period" ORDER BY "period"

SELECT "year","period", AVG("value") AS avg_hourly_earnings
FROM econ.state_industry
WHERE series LIKE 'SMU__000005000000003'
  AND "year" IN (2026)
GROUP BY "year","period" ORDER BY "period"

SELECT "year", industry_code, annual_avg_emplvl, annual_avg_wkly_wage
FROM econ.national_wages
WHERE industry_code IN ('51','5415') AND own_code='5' AND agglvl_code IN ('16','18')
  AND "year" IN ('2019','2020','2021','2022','2023','2024','2025')
ORDER BY industry_code, "year"

SELECT DISTINCT agglvl_code FROM econ.state_wages WHERE industry_code='5415' AND own_code='5'

SELECT "year", SUM(annual_avg_emplvl) AS emp, SUM(total_annual_wages) AS wages
FROM econ.state_wages
WHERE industry_code='5415' AND own_code='5' AND agglvl_code='56'
  AND "year" IN ('2020','2021','2022','2023','2024','2025')
GROUP BY "year" ORDER BY "year"