The $100K H-1B Fee Killed Level 1. Source STEM OPT Seniors Instead.
The Feb 27, 2026 wage-weighted H-1B lottery and $100K fee reshaped international sourcing. Here is the new pipeline math for Q3 2026.
The FY 2027 H-1B cap ran March 4 to 19, 2026 under the new wage-weighted rule, and the selection data is now on the table. Level 4 registrations cleared at roughly 61%. Level 1 cleared at 24.5%. If your 2026 sourcing plan still front-loads OPT new grads out of a top-20 CS program, you are pipelining candidates against a coin flip at 4x the per-hire cost of a senior who is already in the country.
What the Feb 27, 2026 rule actually changed
DHS's final rule replaced the single-entry cap lottery with a weighted one: Level 4 wage offers get four entries in the pool, Level 3 gets three, Level 2 gets two, Level 1 gets one. The tilt is not small. DHS's own projection says Level 1 selection probability drops about 48% while Level 4 rises 107%.
The mechanism matters more than the headline. The wage level is fixed at registration and cross-checked against the Labor Condition Application at petition stage, so a recruiter cannot quietly re-label a Level 1 role as Level 3 without duties, wage floor, and prevailing-wage documentation that survives USCIS scrutiny.
The first-cycle numbers from Manifest Law's FY 2027 dataset:
| Wage level | Selection rate (FY 2027) | DHS-projected change vs old rule |
|---|---|---|
| Level 1 | 24.5% | -48% |
| Level 3 | 69.2% | +55% |
| Level 4 | ~61% | +107% |
Historical H-1B petitions concentrated 83% at Levels 1 and 2. That entire cohort is now negative expected value to source at scale. If your ATS still runs a "new grad international" tag as a first-class pipeline, it is a legacy of pre-2026 economics. DHS also estimates the rule will drive $502 million in first-year wage increases and an $858 million annual wage transfer from Level 1 to higher-level positions, so the reallocation is structural, not a one-cycle blip.
The $100K fee is unstable, but the arbitrage is not
The right sourcing posture is not "wait out the fee." It is "sponsor candidates who are already in the US on OPT, STEM OPT, L-1, or O-1, because the change-of-status carve-out avoids the fee entirely." That has been true regardless of which way the courts have swung.
The timeline in the last twelve months:
- Sept 19, 2025: Proclamation 10973 imposes a $100,000 supplemental fee on new H-1B petitions for beneficiaries outside the US or requiring consular processing.
- Dec 12, 2025: A coalition of state AGs files California et al. v. Trump/Mullin in D. Mass.
- Feb 27, 2026: The wage-weighted lottery rule takes effect for the FY 2027 cap.
- March 4 to 19, 2026: The FY 2027 registration window runs under both rules.
- June 8, 2026: Judge Leo Sorokin vacates the $100K fee as an unlawful tax. Days later, the same court pauses implementation and lets USCIS keep collecting.
- July 24, 2026: The First Circuit denies the government's request to restore the fee during appeal. As of that date, the $100,000 payment requirement is not being enforced.
- Sept 20, 2026: The fee's current form is scheduled to sunset unless extended.
A pipeline built on "the fee is dead" is naive. The appeal is live, the sunset can be extended by proclamation, and any candidate who requires a visa stamp abroad is a live-wire cost risk on a 30-day horizon. The stable move is structural: source people already inside the US whose petitions go through change of status. Under current guidance, F-1 OPT, F-1 STEM OPT, L-1, O-1, and J-1 holders can file change of status without triggering the fee.
This is where sourcing tooling matters. Finding "senior backend engineer, in the US on OPT or STEM OPT for at least 18 months, working in payments or fintech" is not a boolean search LinkedIn can run. It is a plain-English question, which is the exact gap Refolk closes: describe the person and get a ranked shortlist across GitHub, LinkedIn, and the open web.
The senior US bench is roughly 242,000 people
In Refolk's index, there are roughly 242,000 senior, staff, or principal software engineers based in the United States. That is the fee-exempt, change-of-status-eligible bench for anyone who wants to sponsor without paying $100K per petition, and it is the single most important number in the new sourcing math.
The raw dataset:
| Segment | Count | Note |
|---|---|---|
| Senior/Staff/Principal SWEs, United States | ~242,000 | Change-of-status pool, fee-exempt |
| Senior/Staff/Principal SWEs, Canada + India | ~324,000 | Offshore relocation / L-1 pool |
| Software Engineers (all levels), India | ~545,000 | Raw pool the $100K fee taxes |
| Senior US pool as % of India total | ~44% | Fee-exempt bench vs raw India engineer pool |
| Canada+India senior pool vs US senior pool | 1.34x | Offshore is larger, but each new petition carries $100K when fee is in force |
The offshore senior pool is 1.34x larger than the US senior pool, but each new consular petition carries the $100K fee (when in force) plus a Level 3 or 4 wage floor to clear the lottery. Fully-loaded, the smaller US bench wins by a wide margin.
Top regions for the offshore senior pool in Refolk's index are Bengaluru at #1 and Vancouver at #2. Top employers of senior engineers in that offshore pool: EPAM Systems, Google, Meta, GeekyAnts, Nomiso. Those are the natural L-1B intracompany-transfer origins if you already run a Canadian or Indian subsidiary.
Top US-based employers of senior engineers in Refolk's index: Datadog, Starburst, Omada Health, Google. Those are the natural addresses for change-of-status sourcing that never touches the fee at all.
Level 1 is not a coin flip. It is a coin flip against the house.
Why the STEM OPT senior is the real winner
The candidate archetype with the best combined odds is the F-1 STEM OPT holder with two to three years of US experience who has been promoted into a Level 3 wage band. They are inside the country, so no $100K fee. They clear the Level 3 threshold and get three lottery entries against a 69.2% selection rate. They do not need relocation.
This flips the traditional campus-recruiting funnel. Universities are no longer a graduation-year pipeline. They are a five-year pipeline, and the target moment is not offer day. It is the 24-to-36-month mark after OPT starts, when the candidate has:
- Two US performance reviews on record
- Documented duties that map to a Level 3 wage determination
- STEM OPT remaining as a fallback if the lottery misses
- No consular processing requirement, and therefore no fee exposure
The sourcing implication: your 2023 and 2024 new-grad hires at other companies are your 2026 pipeline. A search like "backend engineer, US-based, joined first US job in 2023, currently senior title, working on distributed systems" maps poorly to LinkedIn filters and cleanly to a plain-English question you can ask Refolk.
L-1 and O-1 are the other rational alternatives
If change of status inside the US is Plan A, L-1 intracompany transfers and O-1 extraordinary-ability petitions are Plan B, because neither category triggers the $100K supplemental fee. Any H-1B alternative sourcing strategy for the next 12 months should assume these two channels absorb a meaningful share of what used to flow through the cap.
How the two compare on sourcing signals:
| Channel | Requirement | What to look for |
|---|---|---|
| L-1B intracompany transfer | 1+ year at foreign affiliate in specialized-knowledge role | Current tenure at Bengaluru or Vancouver office of a US-headquartered employer |
| O-1A extraordinary ability | Sustained acclaim in the field | Published research, patents, media citations, judging roles, high-salary evidence |
| Change of status (STEM OPT to H-1B) | Already in US on qualifying visa | US location, US employer history, degree from US STEM program |
Refolk's index shows the L-1 supply concretely: roughly 324,000 senior engineers across Canada plus India, concentrated at EPAM, Google, Meta, GeekyAnts, and Nomiso. If you already operate a subsidiary in either country, that is your legitimate transfer bench. If you do not, it is the pool you would build a Vancouver or Bengaluru satellite around, because standing up a subsidiary for L-1 access has a lower fully-loaded cost than paying $100,000 per H-1B petition at scale.
O-1 sourcing is different. The signal is not location or employer. It is publication record, patent count, conference talks, and salary evidence. That is a query built on open-web signals that GitHub and LinkedIn do not surface cleanly.
The Q3 2026 sourcing plan, in five moves
The actionable rewrite of your international pipeline for the next two quarters comes down to five specific changes:
- Delete "new grad international" as a first-class pipeline. Level 1 clears at 24.5% and your best candidate is competing for a coin-flip slot at $100K downside if the fee gets restored.
- Build a "STEM OPT senior" pipeline off 2023 to 2024 first-US-job hires at other companies. These candidates clear Level 3, are already in the US, and are fee-exempt through change of status.
- Stand up an L-1B pipeline out of Bengaluru and Vancouver. Target EPAM, Google, Meta, GeekyAnts, and Nomiso alumni with 1+ year of qualifying tenure.
- Fund one O-1 sourcing effort per quarter. The pool is small but the fee-exempt status makes each hire disproportionately valuable.
- Track the fee's litigation status weekly. Sept 20, 2026 is the current sunset, the appeal is live, and any consular-processing candidate is a live-wire cost risk until the docket clears.
The through-line: the $100K H-1B fee sourcing environment rewards candidates already inside the US and punishes anyone who needs a visa stamp abroad. The wage-weighted lottery reinforces the same tilt from a different angle, because Level 4 senior engineers get four entries while entry-level participation collapses. Sourcing plans that treat these two rules as separate problems miss that they push in the same direction.
Recruiters who rebuild around that direction will spend the next year hiring from a 242,000-person US bench that most competitors are still ignoring because their tooling still assumes a pre-2026 world.
FAQ
Is the $100,000 H-1B fee still being collected?
As of July 24, 2026, the $100,000 payment requirement is not in effect. The First Circuit denied the federal government's request to temporarily restore the fee during appeal. That said, the appeal is live, the current form of the fee is scheduled to sunset on September 20, 2026 but can be extended by the administration, and the situation has already flipped three times in six weeks. Any sourcing plan that assumes the fee is permanently dead is taking on unnecessary risk. Sourcing candidates who qualify for change of status inside the US remains the stable strategy regardless of how the appeal resolves.
Which candidates avoid the $100K fee entirely?
Workers already in the United States on F-1 OPT, F-1 STEM OPT, L-1, O-1, or J-1 status can file for change of status to H-1B without triggering the $100,000 supplementary fee. The fee applies to new petitions for beneficiaries outside the US or those requiring consular processing. That carve-out is what makes the roughly 242,000-person senior US bench in Refolk's index the highest-leverage pool to source from in 2026.
How much better are Level 4 odds under the new lottery?
In FY 2027 selection data, Level 4 registrations cleared at about 61% and Level 1 cleared at 24.5%, so Level 4 selection was roughly 2.5x Level 1 in realized rates. DHS's projected changes are larger in relative terms: Level 1 odds dropped 48% while Level 4 odds rose 107%. The often-cited "4x" figure refers to lottery entries, since Level 4 gets four entries versus Level 1's one. The realized gap is closer to 2.5 to 3x but still decisive.
What should replace campus recruiting for international hires?
Treat universities as a five-year pipeline rather than a graduation-year pipeline. The target moment is 24 to 36 months after OPT begins, when a candidate has US performance history, has been promoted to a Level 3 wage band, is fee-exempt through change of status, and has STEM OPT remaining as a fallback. Sourcing that cohort means searching by first-US-job date, current title, and location, which is a plain-English query that maps cleanly to a natural-language tool like Refolk.