H-1B Layoff Notices 2020–2026: What the WARN Trend Shows

This is Part 4 of our Layoffs × H-1B series. Earlier parts looked at which sponsors appear in layoff records and what that does and doesn't imply. Here we zoom out to the time dimension: across 2020 through August 2026, how have layoff notices tied to H-1B sponsors moved year to year? The short answer is that there are two very different "big" years, and telling them apart requires understanding how the data was collected.

We track WARN Act notices — the advance-warning filings large employers must submit before mass layoffs — that we could link to companies appearing in H-1B/LCA sponsorship records. This is an August 2026 snapshot covering six tracked states: California, Texas, New Jersey, Illinois, Massachusetts, and Virginia. If you want the underlying counts, see the layoffs section; the methodology page explains how notices are matched to sponsors.

Read these three caveats first

Before any number, three things determine how you should read the trend.

Coverage grew over time. The states did not all enter the dataset at once. Illinois notices begin in April 2020, Massachusetts in April 2021, and California — the largest H-1B state by far — only in June 2026. That means part of any apparent year-over-year increase is more states being watched, not necessarily more layoffs. New Jersey, Texas, and Virginia are the long, consistent series; treat them as the stable backbone and everything else as coverage that widened around them.

Worker counts are floors, not exact totals. 186 notices in the set carry no headcount at all — 175 of those in Illinois. Every "workers named" figure below therefore undercounts reality. Read them as at least this many, never as precise sums.

2026 is a partial year. The snapshot was taken in August, so months are still unreported. And as always in this series: a WARN notice is a layoff-planning filing, not a statement about anyone's immigration status. An LCA is not a visa; certification is not USCIS approval; withdrawn is not denied.

Two "big" years that look nothing alike

The headline is that 2020 and 2023 are both large — but for opposite reasons.

2020 produced the most notices of any year: 139 notices from 88 distinct sponsors. Yet the number of workers named was about 5,607 — roughly 93 workers per counted notice. In other words, 2020 was a year of many small notices. Lots of employers filed, but each filing tended to name a modest number of people.

2023 flips that pattern. It had fewer notices — 129 — but the highest worker count in the series, about 14,095 named, or roughly 145 workers per counted notice. Fewer filings, but bigger ones. That is the year the sponsor-linked layoff footprint peaked in human terms.

Between them sits 2022, the quiet year: just 52 notices and about 4,676 workers named. The dip in 2022 followed by the 2023 surge lines up with the widely reported technology-sector layoff cycle of that period. We note the timing because it is real, but we make no causal claim — this dataset cannot tell you why any employer filed, and we do not attribute layoffs to any single cause.

The plateau after 2023

What is striking is not just the 2023 peak but how the years after it held near that level rather than snapping back to 2022's lows:

After 2023's ~14,095, the count settled to about 13,268 workers in 2024, 12,295 in 2025, and roughly 11,887 already in 2026 across 121 notices — with months still unreported. So 2026 is not a low year; it is a partial year that has already nearly matched full prior years. The gentle downward drift from 2023 to 2025 is modest, and it would be a mistake to read a partial 2026 as a continuation of that decline until the year closes.

One more reason to be careful with the recent end of the series: California only entered the data in June 2026. Because California is the single largest H-1B sponsorship state, its late arrival means the stable backbone (New Jersey, Texas, Virginia) tells you more about genuine change over time than the full-panel totals do. Cross-year comparisons that include newly added states mix real movement with expanding coverage.

Notices vs. workers: why the distinction matters

The 2020-versus-2023 contrast is the single most useful thing in this data, so it is worth stating plainly. If you count notices, 2020 looks like the worst year. If you count workers, 2023 does. Both are true, and they measure different things:

• Notice counts reflect how many separate layoff events were filed. They are sensitive to how many small actions happened and to how many states were being watched.
• Worker counts reflect scale — but they are floors, dragged down by the 186 headcount-free notices.

A reader weighing a job offer, or a journalist writing about "the layoff wave," should decide which question they are actually asking. "How many employers were cutting?" and "How many people were affected?" have different answers here, and the gap between them is the story.

Where sponsor-linked layoffs concentrate

WARN notices are only half of the picture. The other half is who these employers are and where they hire on H-1B in the first place. The sponsors that show up most in layoff records tend to overlap with the largest overall filers, simply because bigger workforces generate bigger notices. For context on the top of that list, this chart shows the leading H-1B sponsors overall:

And because most tracked WARN activity sits in a handful of states, it helps to see where H-1B jobs are filed to begin with:

Keep the coverage caveat in mind when you overlay these: California dominates H-1B filings but only appears in our WARN panel from mid-2026, so its absence from earlier layoff years reflects the data window, not the labor market.

What to take away

The sponsor-linked WARN trend from 2020 to 2026 is best summarized as two distinct waves and a plateau: 2020's many-but-small notices, 2023's fewer-but-larger worker peak, and a sustained high level from 2024 through a still-incomplete 2026. The quiet point was 2022.

But every one of those numbers comes with a footnote. States entered the panel at different times, 186 notices name no workers at all, and 2026 is not finished. Those are not reasons to distrust the data — they are the difference between reading it well and reading it wrong. Use worker counts as floors, lean on the consistent New Jersey–Texas–Virginia series for genuine change, and treat any figure that spans a coverage change as a mix of real movement and wider watching.

Next in the series, we tie the layoff picture back to sponsorship behavior. For the full year-by-year records, browse the layoffs section and the individual 2023 year page.