The $100K H-1B Fee Has a Sunset and a Carve-Out. Source the F-1 Pool.

USCIS sent FY2027 cap selection notices by March 31, 2026, the first full cycle to run with both the $100,000 supplemental fee from Proclamation 10973 and the new wage-weighted lottery live at the same time. Wall Street read the headline and flinched: H-1B LCA filings at top financial firms fell roughly 10% in FY2026 Q1, with Goldman Sachs down more than 60%. Most of those filings did not need to be cut. The fee is a consular-processing fee, and the cohort it doesn't touch is sitting on US soil right now.

What the $100K fee actually says, in one paragraph

Proclamation 10973 took effect September 21, 2025 and imposes a $100,000 supplemental payment on certain H-1B petitions. USCIS has since clarified that it does not apply to petitions for workers already in the US whose employer requests a change of status. Even better for planning purposes: an approved COS beneficiary stays exempt if they later leave the country and pick up a visa stamp at a consulate. The proclamation has a roughly 12-month built-in sunset around September 21, 2026, and is already in litigation on APA, unlawful-tax, and congressional-fee-authority grounds. Translation: FY2027 cap season is the only full cycle this fee is guaranteed to run, and the carve-out is wider than the rule.

USCIS has explicitly confirmed that F-1 OPT students eligible for change of status to H-1B in calendar year 2026 are not affected by the proclamation. The legal community's framing is blunter: the majority of first-time H-1B petitions in each annual lottery are for students, so the proclamation will not have as large an impact as its plain language originally suggested.

60%

Who actually pays $100K, and who doesn't

Pay the fee:

• New H-1B petitions where the beneficiary is outside the US and will be issued a visa at a consulate.
• Cap-subject petitions for offshore hires brought in for the first time.

Skip the fee:

• Change-of-status petitions for F-1 OPT and STEM-OPT workers already inside the US.
• Cap-exempt employers (universities, affiliated nonprofits, government research orgs).
• Approved COS beneficiaries who later travel and consular-stamp.
• L-1 to H-1B COS, O-1 to H-1B COS, and similar in-country status changes.

National-interest exemptions exist on paper. DHS has called them extremely rare and routes requests to H1BExceptions@hq.dhs.gov. Don't plan around them.

Pre-proclamation, an H-1B hire cost roughly $2,010 to $6,185 in government fees. Add $100K and a single offshore hire exceeds what most mid-sized employers can justify. That math is real. It just doesn't apply to the candidate you should have been targeting anyway.

Why wage-weighted selection makes the F-1 pool more valuable, not less

The new selection process took effect February 27, 2026. Level 1 gets 1 entry, Level 2 gets 2, Level 3 gets 3, Level 4 gets 4. The popular reading is that this hurts foreign grads. For tech, that read is backwards.

F-1 grads from CMU, UIUC, Georgia Tech, UT Austin, Berkeley, and Stanford going into FAANG, quant, and AI roles are routinely Level III at offer and Level IV within eighteen months. They are already inside the country, COS-eligible, and exempt from the $100K. They also get three to four lottery entries each. Stack those facts:

1. No supplemental fee on COS.
2. 3x to 4x selection odds under the new weighting.
3. Cap-gap work authorization once you file, which extends OPT until USCIS decides.

Capital One filed about 4% more this cycle, with roles concentrated in data science, ML, and senior data engineering. That's not luck. That's a wage architecture that crosses Level III on the relevant SOC codes. Nvidia was the lone major tech filer that did not cut. Goldman cut 60%. Two of those three companies are pricing this correctly.

The fee is a sunset clause, not a regime. Sponsoring engineers in 2026 is a wage-architecture problem, not a legal one. </pull> Wait, the renderer wants fenced blocks. Let me fix that.

pull The fee is a sunset clause, not a regime. Sponsoring engineers in 2026 is a wage-architecture problem, not a legal one.

## The arbitrage window and why it closes

Total H-1B registrations dropped to 358,737 in FY2026 from 479,953 in FY2025 and a peak of 780,884 in FY2024. Demand at the registration door fell by more than half in two years. Some of that is bot-cleanup from the prior beneficiary-centric reform. A meaningful chunk is employers misreading the $100K fee as universal and quietly walking away from candidates they could have sponsored cleanly.

JPMorgan Chase fell from 724 filings in Oct-Dec 2024 to 516 a year later, losing its #1 spot. That's nearly 29% in a single quarter at the most sophisticated immigration team in financial services. If JPMorgan is overcorrecting, smaller employers are overcorrecting harder. The arbitrage is six to twelve months: long enough to staff a team, short enough that you have to start now.

Where the F-1 pool actually lives

You will not find this cohort with a LinkedIn title search. Visa status isn't a profile field, and "F-1" is not a skill. The signals are oblique:

• A .edu email on a US university paired with a US-based employer in the last 18 months.
• GitHub commit history that starts during a US graduate program and continues at a US company without an international relocation.
• Career-services networks at top CS programs (CPT and OPT placement lists, often public).
• STEM-OPT employer rosters visible through E-Verify.
• On-campus AI and ML clubs, especially the ones with active GitHub orgs.
• User bases of immigration-tooling vendors like Alma, Rippling, and Deel.

Stitching those signals together is exactly the kind of query that breaks Boolean. It's also why we built Refolk: you describe the person in plain English ("US-based F-1 OPT backend engineers from a top-25 CS program, 2 to 5 years experience, currently at a Series B or later startup") and get a ranked shortlist across GitHub, LinkedIn, and the open web. Visa-status inference comes from the pattern, not a checkbox.

What to actually do before September 2026

The proclamation sunsets around September 21, 2026 unless renewed. Litigation could end it earlier. Either way, FY2027 is the cycle to act on.

1. Re-rank your existing pipeline by COS eligibility

Pull every candidate currently in your ATS who would need sponsorship. Tag the ones already inside the US on F-1, OPT, STEM-OPT, L-1, or O-1. Those are your zero-fee hires. The offshore tier is where the $100K math actually bites, and for most roles you can backfill from the COS tier.

2. Map comp to wage levels before you make offers

If multiple employers register the same beneficiary, the lowest wage level among them controls, and USCIS can deny or revoke for inflated wages or post-registration wage cuts. That means comp band design now decides lottery odds. Pricing a senior backend role at Level II to save $30K of base pay costs you 50% of the lottery entries. Talent and finance need to be in the same room before you publish the JD.

3. Source the cap-gap cohort deliberately

F-1 OPT expiring after April 1, 2026 plus a cap-subject H-1B COS filing equals automatic work-authorization extension until USCIS decides. Under wage-weighted selection, more Level III and IV F-1s will get picked, which means more candidates will need cap-gap. OPT expiration timing is now a useful sourcing filter, not a paperwork detail. Refolk lets you ask for it directly in the query rather than reverse-engineering it from graduation year.

4. Don't burn the consular pool, just price it correctly

If you have an offshore senior staff or principal candidate whose total cost of ownership is multi-million-dollar over four years, $100K is a rounding error. The fee kills the median junior offshore hire, not the strategic one. Decide which is which before you cut your filings 60%.

5. Build the FY2028 bench now

The fee may sunset before you need it. The candidates won't. The F-1 grads finishing in May 2026 are the ones you sponsor in March 2027 for FY2028 caps, and the strongest ones already have multiple offers by January. Refolk queries against current PhD candidates and recent grads at named programs are how engineering leaders are quietly pre-staffing the next cycle while their competitors are still arguing about the fee in Slack.

358,737

The headline most companies got wrong

The headline was "H-1B just got $100K more expensive." The accurate version is "H-1B just got $100K more expensive for offshore consular hires for one cycle, while the wage-weighted lottery made the in-country F-1 pool 3-4x more likely to win the lottery and exempt from the fee."

The companies that read the first headline cut filings. The companies that read the second one are quietly running the table on Level III and IV F-1 engineers their competitors stopped pursuing. You have until roughly September 21, 2026 before the policy changes again, and probably six to twelve months before larger employers recalibrate. That is enough time to build a team. It is not enough time to wait for clarity.

FAQ

Does the $100K fee apply to H-1B extensions or transfers?

No. The proclamation targets new petitions tied to consular visa issuance. Extensions, amendments, and H-1B portability transfers for workers already on H-1B status in the US are not subject to the supplemental fee. USCIS has been consistent on this since the September 2025 effective date. The cohort that pays is offshore new hires entering for the first time on a consular stamp.

Are F-1 STEM-OPT students automatically exempt, or does it depend on timing?

USCIS has confirmed F-1 OPT students eligible for change of status to H-1B in calendar year 2026 are not affected. The mechanic that matters is filing for change of status while the candidate is in valid F-1 or OPT status inside the US. If the candidate departs and seeks a consular stamp before USCIS approves the COS, the analysis gets messier. Approved COS beneficiaries who later travel and stamp remain exempt.

How does the wage-weighted lottery actually change selection odds?

Level 1 candidates get 1 lottery entry, Level 2 gets 2, Level 3 gets 3, Level 4 gets 4. A Level IV senior engineer is four times more likely to be selected than a Level I new grad in the same registration pool. Because the F-1 OPT pipeline at top CS programs feeds disproportionately into Level III and IV roles at FAANG, quant, and frontier AI companies, the weighting structurally favors the cohort that is also exempt from the $100K fee.

What is the fastest way to find F-1 OPT engineers without a visa-status field on LinkedIn?

You triangulate. US graduate program plus continuous US employment plus a graduation date in the last three years is the rough heuristic. The signals that work in practice are GitHub commit timelines, .edu email histories, university career-services placements, and STEM-OPT employer rosters. Doing this by hand is slow. Tools like Refolk let you describe the candidate profile in plain English and get a ranked shortlist that infers the pattern from public data across GitHub, LinkedIn, and the open web.