Stripe 46, OpenAI 27: The H-1B Target List After the $100K Fee

If you are on H-1B and watching your target-employer list evaporate, the September 2025 $100,000 fee did not shut the door. It split the door in two. On one side, a handful of cash-rich AI labs kept filing at record pace. On the other, everyone else quietly stopped writing $100K checks and started sourcing candidates who were already in the country.

USCIS data through the first half of fiscal 2026 makes the split concrete. OpenAI landed 27 new H-1B approvals, already past its full-year FY2025 total of 21. Stripe hit 46, double its 23 from a year earlier. Meanwhile Maxine Bayley, an immigration partner at Duane Morris, reports zero clients have paid the fee. That is the map you need to rebuild your target list against.

Who is actually still filing new H-1Bs in FY2026

OpenAI, Stripe, Anthropic (barely), Nvidia, and the big consulting sponsors. Almost everyone else is either pivoting to transfers or pulling back entirely.

The Pitchbook cut of USCIS approvals through June 30, 2026 is the cleanest signal available. OpenAI had 304 total H-1B petitions approved by mid-FY2026, up from 211 across all of FY2025, jumping to 79th on the sponsor rankings after not cracking the top 100 the year prior. Stripe doubled its new-approval count in six months. Certified LCA filings tell the same story on the front end: Anthropic went from 10 to 59 (about +490%), OpenAI 20 to 63 (+215%), Nvidia grew 19% off a much larger base of roughly 765.

The pullback list is just as informative. Amazon, Google, Microsoft, Meta, and Apple all posted steep or moderate declines. That is not a shrinking market. It is a redistribution from platform generalists at FAANG toward frontier-AI infrastructure at labs that will pay the fee out of pocket.

Why transfers matter more than new sponsorships

Transfers do not trigger the $100K fee, so for any candidate already in the US on H-1B, the target list should be sorted by transfer volume rather than sponsorship rankings. This is the single most important recalibration to make right now.

The USCIS numbers include both first-time approvals (which trigger the fee) and sponsorships for workers already stateside (which do not). OpenAI's 126 transfers dwarfed its 27 new approvals. Databricks pulled in 89. Rippling took 54. Anthropic absorbed 48 while filing only one new-from-abroad petition. Read those numbers again with a candidate's eye: if you are already on H-1B, you are not competing against people in Bangalore for a $100K-fee slot. You are competing against other in-country transfers for a no-fee slot, and the employer wants you specifically because you cost them nothing extra.

The resume implication is direct. Your header, summary, and location line should telegraph "already in the US, transfer-ready, no fee owed" in the first three seconds of reading. That is not a vibe. It is a specific rewrite:

• Location as a US metro, not "Open to relocation"
• Current-status line: "Currently on H-1B, cap-exempt petition transferable" or "H-1B, no cap-subject filing required"
• Employment dates unbroken to the month, with any consulting or contract work explicitly bridging gaps
• A one-line note in the summary that names the transfer path ("Portable H-1B, priority date current")

Rewriting a resume five different ways for OpenAI, Databricks, Stripe, Rippling, and Anthropic is the grind that used to eat a full weekend. That is the exact work Refolk takes off your desk: paste the posting, get your own history rewritten against it, with the transfer-ready framing already surfaced where an ATS parser and a recruiter both catch it.

The 45-day gap rule is quietly killing candidates

Per CDF Labor Law's read of the USCIS clarification, an employer avoids the $100K fee only if the H-1B candidate has been unemployed fewer than 45 days when the petition is filed. That is a hard fire-sale window with a hard edge.

Combine that with the 60-day grace period H-1B workers get after losing a job, and the math is brutal. You have roughly 45 days from your last day of work to be filed for, not just interviewing. Miss that and you become a $100K liability the moment an employer runs your case. This is why gap-management on the resume matters more than any summary rewrite:

1. Never leave a month unaccounted for. Consulting, part-time, or contract work goes on the resume with clear dates, even if it was two weeks.
2. Concurrent cap-exempt work counts. A university-affiliated or nonprofit research role in parallel keeps status intact and does not trigger the fee.
3. Don't round dates to years. "2024 to 2025" reads as up to 24 months of ambiguity. Use "Mar 2024 to Feb 2025" so a recruiter can verify you are inside the 45-day window without asking.
4. Move the last-day date up front. If you are within window, say so. "Available immediately, last employed [date]" belongs in your summary line, not buried in a cover letter.

The cap-exempt front door, not backup plan

Cap-exempt petitions (filed by universities, university-affiliated nonprofits, government research organizations, and qualifying research institutions) are not subject to the annual lottery and are generally not subject to the $100,000 fee. That is a permanent structural advantage, not a fallback.

For FY2027, USCIS received roughly 480,000 registrations for 85,000 regular cap slots, a selection rate around 18%. Meanwhile the wage-weighted lottery that took effect February 27, 2026 skews selection toward higher-paid registrations, pushing mid-band candidates further out of the running. Cap-exempt is not the consolation prize anymore. For a lot of profiles it is the shortest path to a valid H-1B.

The named cap-exempt employers worth building a target list around:

• Research institutes: Broad Institute, Howard Hughes Medical Institute, Salk Institute for Biological Studies, Scripps Research Institute, Fred Hutchinson Cancer Center, Battelle Memorial Institute
• University-affiliated medical systems: Johns Hopkins Medicine, UCSF Medical Center, Mayo Clinic
• University research computing: any R1 university's central IT, HPC, or genomics core

There is also a rarely-used move worth knowing: you can hold concurrent H-1Bs, one cap-exempt and one cap-subject, which means a research-institute role can anchor your status while you interview at a Stripe or an OpenAI without a fee ever coming due.

The FAANG-to-frontier resume pivot

A resume optimized for platform generalist work at Amazon, Google, or Microsoft in 2023 reads as the wrong shape in 2026, because the sponsors still writing checks are frontier-AI labs hiring for a narrower and deeper stack. Rewrite for the stack the hiring queue actually wants.

The delta shows up in what recruiters at Nvidia, OpenAI, and Anthropic scan for first:

• Distributed training at scale (not just "trained models"): named frameworks, cluster size, throughput numbers
• Inference optimization: quantization, KV-cache, speculative decoding, batching
• CUDA and kernel work where relevant, or Triton/ROCm equivalents
• Eval infrastructure: not benchmark scores, but the harness you built to produce them
• RLHF / post-training pipelines: reward modeling, preference data curation, PPO/DPO in production

A resume that led with "designed microservices, mentored junior engineers, drove roadmap" reads as expensive filler to a lab that needs the third person on the inference team. Rewriting your existing bullets to lead with the numbers a frontier-lab recruiter looks for is mechanical work, and Refolk handles it against each posting: same underlying history, different lead bullets for a Databricks Spark role vs an Anthropic inference role.

The consulting fallback is still real. Deloitte Consulting filed 3,209 LCAs at an average salary of $117,683 for advisory and technology roles. Cognizant sponsored 2,493 approvals averaging $131,440 for digital transformation and IT work. Those are not glamour placements, but they are volume sponsors that keep filing, and for candidates outside the AI-infra shortlist they remain the highest-probability route to a valid H-1B in 2026.

The clock: what changes on September 24, 2026

The current playbook has an expiration date. On August 27, 2026 the White House's Office of Information and Regulatory Affairs cleared a DHS proposal that would end the H-1B grace period for workers who lose their jobs, with public comments due September 24, 2026. Separately, DHS proposed a $103,265 fee on cap-subject petitions that would generate an estimated $8.8 billion annually against 85,000 cap-subject filings.

If the proposals finalize as drafted:

• The grace period disappears, meaning any layoff becomes a same-week emergency
• The transfer loophole tightens as the $103K fee expands the fee's reach
• Databricks-style absorption stops working as a fee-free strategy
• The 45-day filing window collapses toward zero

None of that has happened yet. A federal judge already struck the September 2025 $100K fee down as an unlawful tax, and the $103K rulemaking is the administration's regulatory workaround. But the window in which "transfer-ready, in the US, under 45 days out" is a viable resume angle is measured in months, not years. Get the target list rebuilt now, get the resume rewritten now, and get the applications out before the comment period closes.

The candidates who come out of this cleanest will be the ones who stopped applying to Amazon and Google out of habit, built a 15-employer list that mixes OpenAI-tier fee-payers, transfer-heavy destinations like Databricks and Rippling, cap-exempt research institutes, and volume consulting sponsors, then rewrote a version of their resume for each.

FAQ

Which companies are still paying the $100K H-1B fee?

Based on USCIS approval data through June 30, 2026, the clearest fee-payers are OpenAI (27 new approvals in H1 FY2026), Stripe (46), and Nvidia (approximately 765 certified LCAs, up 19% year over year). Anthropic has almost entirely pivoted away from new-from-abroad hiring (one approval) toward transfers (48 absorbed). Amazon, Google, Microsoft, Meta, and Apple have all posted steep or moderate declines. Mid-market and startup sponsors have effectively stopped writing $100K checks, per Maxine Bayley of Duane Morris, who reports zero clients have paid the fee.

Do H-1B transfers require the $100K fee?

No. The $100,000 Presidential Proclamation fee introduced in September 2025 applies to new cap-subject petitions, meaning first-time filings for workers coming from abroad. Sponsorships for candidates already on H-1B in the US are transfers, and USCIS guidance confirms they are not fee-triggering, provided the candidate has been unemployed fewer than 45 days when the petition is filed. That distinction is why OpenAI took 126 transfers versus 27 new approvals in the same six-month period, and why your resume should lead with "already in the US, transfer-ready" language.

Are cap-exempt employers still a good target in 2026?

Yes, and arguably more than ever. Cap-exempt petitions are not subject to the annual lottery (roughly 18% selection rate for FY2027) and are generally not subject to the $100K fee. Named targets include the Broad Institute, Howard Hughes Medical Institute, Salk Institute, Scripps Research, Fred Hutchinson Cancer Center, Battelle Memorial Institute, plus university-affiliated systems like Johns Hopkins Medicine, UCSF Medical Center, and Mayo Clinic. The tradeoff is that more candidates are funneling toward these employers as the cap-subject lottery tightens, so per-opening competition has risen.

How should I rewrite my resume for the transfer path?

Front-load the transfer-ready signal: US metro location, current H-1B status line in the summary, unbroken employment dates to the month, and a lead bullet stack that matches what the fee-paying employers actually hire for (distributed training, inference optimization, eval infrastructure for AI labs; payments infrastructure for Stripe; data platform work for Databricks). Kill any gap longer than 45 days by filling it with named consulting or contract work. Then rewrite once per posting, because a resume that reads as generic to an OpenAI recruiter reads as expensive filler.