The $100K Fee Killed Level I H-1B. 3,002 Level IV Names Remain.
The FY2027 wage-weighted H-1B lottery and $100K fee bifurcated the US technical talent pool. Here is what to source instead, and when.
If you run technical hiring in the US, the sourceable pool you had in January is not the pool you have now. The FY2027 H-1B cycle just closed under a wage-weighted lottery where Level IV petitions get four times the entries of Level I, and Proclamation 10973's $100,000 fee - currently vacated but with a Sept 20, 2026 expiration and a $103,265 regulatory replacement in the pipeline - has made overseas entry-level sponsorship indefensible on any hiring plan longer than three weeks. Entry-level sourcing moves onshore to OPT, and a bidding war concentrates on roughly 3,000 Level IV names.
What actually changed in the FY2027 H-1B cycle
USCIS replaced the random lottery with a wage-weighted selection, and a $100,000 per-petition supplemental fee attached to any H-1B beneficiary outside the US. Together they made Level I sponsorship of overseas candidates a coin flip nobody rational will pay for.
The wage-weighted final rule ("Weighted Selection Process for Registrants and Petitioners Seeking to File Cap-Subject H-1B Petitions") was issued by DHS on Dec 23, 2025, published in the Federal Register Dec 29, 2025, and became effective Feb 27, 2026, in time for the FY2027 registration window. Each beneficiary is entered into the selection pool a number of times based on the OEWS wage level of the offered job:
- Level I (entry): 1 entry
- Level II: 2 entries
- Level III: 3 entries
- Level IV (fully competent): 4 entries
If the same beneficiary is registered at different wage levels by different sponsors, USCIS uses the lowest level for weighting. The old "one shop registers you at L1, another at L4" arbitrage is dead.
Layered on top: Proclamation 10973, effective 12:01 a.m. EDT Sept 21, 2025, imposed a $100,000 supplemental fee on new H-1B petitions for beneficiaries outside the US. The fee does not apply to change of status, extensions, or amendments for people already in the country. On July 24, 2026, the First Circuit denied the government's motion to stay the district court's vacatur, so the fee is not currently collectable. But the proclamation's stated expiration is Sept 20, 2026, and DHS has a replacement $103,265 fee moving through notice-and-comment.
Why Level I sponsorship is economically dead
Level I sponsorship of an overseas hire is a $5,000 to $8,000 legal bet on a roughly 1-in-7 outcome, against a $100,000 fee (or its regulatory successor) if you win. No rational hiring plan pencils out.
Historically, all H-1B registrations averaged a 29.59% selection rate across FY2021 to FY2025. Under the new weighting, that pooled average splits: Level IV climbs materially above it, and Level I collapses well below.
Do the math on a new grad SWE offer at OEWS Level I (typically $70,000 to $80,000 base in most metros):
- Base salary at L1: ~$75,000
- Supplemental fee if collectable: $100,000 (1.3x to 1.5x annual salary)
- Legal and filing costs: $5,000 to $8,000
- Selection probability: sharply below the 29.59% pooled historical average
The rational response is the one big employers are quietly executing: hire the same candidate on OPT for three years, and only file H-1B once they have been promoted into a Level III or IV wage band.
The OPT pool nobody has priced in
The domestic technical talent pool absorbed the shock: 418,781 people held OPT authorization in CY2024, and roughly 205,000 of them are CS or engineering graduates already inside the US. These candidates dodge the $100K fee entirely, because change-of-status petitions for people already present are exempt.
Optional Practical Training (OPT) is a 12-month work authorization tied to an F-1 student visa, with a 24-month STEM extension that brings total runway to 36 months. Per NAFSA and CRS analysis of DHS data:
- 418,781 total OPT authorizations in CY2024
- 31% majored in computer science (~130,000)
- 18% majored in engineering (~75,000)
- 165,000+ STEM-OPT authorizations in 2024, surpassing pre-pandemic highs
- India F-1 records: 422,335 (+11.8% YoY)
- Nepal STEM OPT: 3,752 participants (+42% enrollment growth)
Top STEM-OPT employers per SEVP include University of California, Walmart, Goldman Sachs, Tesla, and Intel. These are the templates: hire on OPT, defer H-1B to a promotion cycle when the wage band lands at Level III or higher.
A CS grad you hire on OPT in mid-2026 has work authorization through mid-2029. The $100,000 fee proclamation expires before their first OPT year ends. The wage-weighted lottery only matters when you eventually file, and by then they are at L2 or L3, not L1. You have converted a coin flip at $100K into a certainty at zero fee.
The friction is that OPT candidates do not sit in a clean database. They cluster on campus career portals, in day1cpt.org threads, on Interstride and Interex Group, and in university Slack groups. Boolean searches on LinkedIn miss most of them because they list "New Grad" or "MS CS 2026" instead of the exact stack you want. This is where Refolk helps: describe the person in plain English ("MS CS graduating spring 2026 in the US with PyTorch and CUDA on their GitHub, open to relocation") and get a ranked shortlist pulled from LinkedIn, GitHub, and the open web, which is where OPT candidates actually publish.
The Level IV bidding war: 3,002 names
At the top end, the wage-weighted rule concentrates competition on a tiny cohort. In Refolk's index of professional profiles, there are 3,002 US-based ML/AI engineers with PyTorch on their profile. That is the Level III/IV pool every FAANG will petition for FY2028.
The top current employers of that cohort in Refolk's index skew exactly where you would expect: Meta, Apple, TwelveLabs, fileAI, and AppFolio. Bay Area, Seattle, and NYC absorb most of the concentration. With Level IV registrations getting 4x weight and the lowest-level tiebreaker closing the arbitrage, every large sponsor will file on the same names. Expect offered wages to inflate 15% to 25% to clear Level IV OEWS thresholds, and retention packages at incumbents to stiffen to keep the L4 pool from leaking to the next tier of AI labs.
Level I sponsorship is dead. Level IV sponsorship is a bidding war over 3,002 people. The interesting hiring happens in between, on OPT.
The comparable numbers, on one page
| Cohort | Figure | Source |
|---|---|---|
| US-based ML/AI engineers with PyTorch | 3,002 | Refolk index |
| Total OPT authorizations, CY2024 | 418,781 | NAFSA / CRS / DHS |
| STEM-OPT authorized, CY2024 | 165,000+ | ApplyBoard / SEVP |
| CS majors in 2024 OPT pool (31%) | ~130,000 | NAFSA / CRS |
| Historical H-1B selection rate FY21 to FY25 | 29.59% | DHS via Dworsky Law |
| Level IV vs Level I lottery weight | 4:1 | DHS final rule matrix |
| $100K fee vs Level I annual wage | 1.3x to 1.5x | OEWS L1 SWE wages ~$70 to $80K |
The Sept 20, 2026 cliff and why it does not save you
The $100,000 fee is not collectable today, but building a hiring plan on that assumption is a three-week bet, not a strategy. The proclamation expires Sept 20, 2026 unless reissued, and DHS has a $103,265 regulatory replacement drafted.
Three scenarios sit on the table:
- Proclamation lapses, no replacement. The fee disappears entirely. Wage-weighted lottery remains. Level I sponsorship is still economically weak because of the entry-weight math, but overseas sponsorship at Level III or IV becomes viable again.
- Proclamation is extended or reissued. The fee snaps back for beneficiaries outside the US. OPT hiring stays the dominant entry-level path.
- Regulatory $103,265 fee finalizes. Roughly the same economics as scenario 2, with a harder-to-litigate legal footing because it went through notice-and-comment.
The change-of-status carve-out survives every scenario. Hiring someone already in the US on OPT is fee-exempt regardless of what happens to the proclamation. That is why the tactical answer, not the political one, is to move now.
What to actually do in the next 90 days
Rebuild your sourcing plan around two pools: OPT candidates already inside the US for anything below Level III, and the ~3,000 Level IV names for anything senior. Stop pretending the middle path still works.
1. Rewrite entry-level reqs against the OPT pool
Anything you would have sponsored at L1 or L2 becomes an OPT hire. Practical moves:
- Target Fall 2025 and Spring 2026 MS CS and MS ECE cohorts at top-30 US programs
- Screen for STEM-designated degrees (CIP codes on the DHS list) to qualify for the 24-month extension
- Prioritize candidates who have already done a summer internship on CPT, because they are one paperwork step from full-time
- Build a 3-year retention model, not a 6-year one, and plan the H-1B filing for their L3 promotion window
2. Map the 3,002 Level IV names before FY2028 registration opens
March 2027 registration is closer than it feels. If you sponsor at Level IV, you are competing against Meta and Apple for a pool small enough to enumerate. This is the exact gap Refolk closes: describe the Level IV profile you want ("ML engineer with PyTorch and distributed training experience, currently in the US, staff or principal level, based in SF Bay or Seattle") and get the ranked shortlist instead of scraping LinkedIn for a week.
3. Poach with the fee math in mind
The change-of-status exemption follows the person, not the employer. If you poach an OPT holder from Tesla to your company, that is a fresh cap-subject petition and the fee applies (in scenarios 2 and 3 above). Poaching someone already on H-1B is an amendment, and the fee does not apply. The counterintuitive move: poach H-1B holders aggressively before Sept 20, 2026 to lock in fee-exempt amendments, and hire OPT holders directly from campus to avoid the poaching penalty entirely.
4. Kill the offshore sponsorship pipeline for anything under Level III
If a partner in Bangalore was feeding L1 and L2 sponsorship candidates into the US pipeline, that pipeline is a $100,000-per-head bet on a weight-1 entry. Redirect the budget into US campus recruiting and OPT sourcing. The OPT pool publishes on GitHub and personal sites long before it shows up on LinkedIn, which is where cross-source search beats single-site Boolean.
FAQ
Is the $100,000 H-1B fee actually being collected right now?
No. As of July 24, 2026, the First Circuit denied the government's motion to stay the district court's vacatur of the policy implementing Proclamation 10973's $100,000 fee. USCIS is not collecting the fee while the appeal proceeds. The proclamation itself is scheduled to expire Sept 20, 2026 unless extended or reissued, and DHS has a separate $103,265 regulatory fee moving through the Federal Register. Any hiring plan should assume some version of the fee returns.
How does the wage-weighted lottery actually work?
USCIS enters each H-1B beneficiary into the selection pool a number of times equal to the OEWS wage level of the offered position: 1 entry for Level I, 2 for Level II, 3 for Level III, 4 for Level IV. If the same person is registered by multiple sponsors at different levels, the lowest level is used. The historical pooled selection rate was 29.59% across FY2021 to FY2025; under weighting, Level IV rises materially above that baseline and Level I falls materially below it.
Why is OPT the answer instead of just filing H-1B at Level III?
You can file at Level III, but only if the candidate legitimately qualifies for that wage band and the job duties support it. For a new grad with no prior US experience, L1 is what the labor certification will support, and misclassifying to L3 invites a DOL audit. OPT lets you hire the same person immediately, get them 12 to 36 months of work authorization at zero fee exposure, promote them into a legitimate L3 wage band, and file H-1B under favorable weighting when they are ready.
Where do you actually find OPT candidates at scale?
Not on LinkedIn Recruiter alone. OPT candidates cluster on university career portals, day1cpt.org, Interstride, Interex Group, and their personal GitHub and university-hosted pages. The signal you want (STEM CIP code, graduation date, GPA, tech stack) sits across LinkedIn, GitHub, and the open web, which is why plain-English sourcing across all three sources beats Boolean strings on any single site.
Try it on the search you came here for
Stop building boolean strings. Just describe the person.
Type one sentence. I plan the search, read GitHub, public LinkedIn and Crunchbase records, and the open web as it is right now, and hand back a ranked list with the reason next to every name.
01Describe them
One plain sentence. Role, city, stack, stage, whatever matters to you.
02I read the web live
GitHub, public LinkedIn and Crunchbase records, the open web. Not a database that went stale last quarter.
03You read the shortlist
Ranked, with the reasoning under every name. Open a profile, ask a follow-up, narrow it down.
- Staff backend engineers in NYC who shipped Rust in production
- Series A fintechs in SF under 50 people, growing headcount this year
- Maintainers of fast-growing Rust web frameworks on GitHub
- No boolean, no filters, no seat to buy. One box.
- Read at search time, so a profile updated yesterday counts today.
- Every step visible as it runs, every name with its reason.
500 free credits on sign-up. No card, no demo call. See real searches.