The Cap-Exempt Advantage.

Every year thousands of international researchers register for the H-1B cap, lose, and conclude the United States is closed to them. A large share of those people were never obliged to be in that queue.

Institutions of higher education, nonprofits affiliated with them, and nonprofit or governmental research organisations are exempt from the numerical cap. A petition filed by an exempt employer is not entered in the lottery, is not counted against the cap, and can be filed on any day of the year, up to six months before the start date. No March registration window. No April result. No October start.

Figures are government filing fees only and vary with employer size and premium processing. Attorney or preparer costs are separate.

The report has two independent halves and keeping them apart is the whole design. The position sweep says who is hiring right now. The filing data says who has actually sponsored. Neither substitutes for the other.

Trap one: the column count moves between quarters

In FY2025 the Q1 file carried 97 columns and Q2 through Q4 carried 98. Anything reading columns by position shifts every wage field by one for three quarters out of four and produces medians that look entirely plausible and are wrong. Every field here is extracted by column name. This is not hypothetical: an earlier cut of this analysis published a national postdoc median of $66,492 on a sample of 6,449. Corrected, it is $67,830 on 5,470.

Trap two: rows are worksites, not petitions

A single petition covering four locations appears four times. Raw row counts overstate the market roughly fourfold. Everything above is deduplicated on case number before anything is counted.

Verified open research and faculty positions at cap-exempt US institutions on 26 August 2026, sorted into thirteen fields. Coverage spanned HERC, AcademicJobsOnline, UC Recruit, Interfolio mirrors, Chronicle of Higher Education, Inside Higher Ed, Nature Careers, Science Careers, discipline society boards, USAJOBS and institutional applicant tracking systems.

Every one of these positions is listed individually in section 04, with a direct application link. The per-field reports and spreadsheets are in section 05.

Every position from all thirteen spreadsheets, in one place. Pick a field, filter by role, tier or state, search by institution or keyword, then apply directly. Nothing here routes through us.

Each field has its own thirteen-page report and a spreadsheet listing every open position we verified, with a direct application link on each row. Both are free to read and free to share, and both open in Google Drive.

This is the most expensive mistake in an academic job search and almost nobody knows about it. A large share of US institutions never use the word "postdoc" on a requisition. They hire the identical role under a different title, and every search run for "postdoc" returns nothing from them.

Every cap-exempt institution could sponsor. Far fewer demonstrably do. When a department says "we only do J-1", the filing record is the evidence that this is one administrator's habit rather than an institutional policy.

Field-level filing counts look alarmingly thin, and that is a classification artifact rather than a market signal

Pharmacy and pharmaceutical sciences is the clearest example. Note that this month's position figure is a verification sample rather than a sweep, so the two numbers below are not comparable with each other; the petition count is a full federal year. Only 126 certified petitions across all US academic employers were coded to that field in FY2025, against 38 live open positions in the same field on the sweep date. The gap is not a contradiction. Researchers in this field are routinely filed under adjacent occupational codes, most often biology, chemistry or general postsecondary teaching, so their petitions land in another field's column.

Wages here are the offered rate, annualised from the stated unit of pay. Where an employer filed a range, the low end is used, so every figure below is conservative.

This section is the one most likely to be out of date in any advice a candidate reads elsewhere, including this page. Verify it against the dockets before anyone files.

Twice in twelve months Washington has tried to put a six-figure price on hiring an H-1B worker. The first attempt, a $100,000 proclamation signed last September, reached people being hired from abroad, carried no university carve-out, and has spent this year losing in court. The second arrived on 25 August as a formal DHS proposal: $103,265 on every cap-subject petition, with institutions of higher education, their affiliated nonprofits and nonprofit research organisations explicitly excluded. One measure threatened the exempt sector; the other widens its advantage.

The two are easy to confuse because the amounts are close, and confusing them leads to bad decisions in both directions. The timeline below carries both stories in order; the table after it pulls them apart.

What the $103,265 number actually is

The proposed figure is not an estimate of what it costs to adjudicate a petition. DHS identified roughly $8.8 billion in costs to recover across the agencies that administer the immigration system, divided it by an assumed 85,000 cap-subject petitions a year, and arrived at $103,264.57. That is a revenue target divided by a headcount, and it is worth understanding plainly, because it is the part most likely to be contested during the comment period.

It applies to cap-subject petitions only, at the point of filing, across both the regular cap and the advanced-degree exemption. It does not reach petitions filed by cap-exempt employers under INA 214(g)(5)(A)-(B), and it does not reach extensions, amendments, or transfers for someone already counted against the cap. Which puts the advantage this report documents into a single row of arithmetic:

Three readers, three different consequences

Everything above assumes an institution files on your behalf. For a large share of researchers that assumption is unnecessary. If your record already carries publications, citations, invited talks, grant participation or journal review work, the strongest routes are the ones you control yourself.

A certified labor condition application is a wage attestation filed before the I-129. It is not a job offer and not an approved petition, so it overstates approvals. Read the filing tables as "this institution has sponsored and its HR office knows how", never as a vacancy list. The position sweep is the vacancy list.

• Portals that block automated reading are absent despite being active hirers. Their absence means nothing. Check them in a browser.
• Interfolio detail pages could not be read. Where a posting routes through it, salary and visa fields are missing, not absent. A blank sponsorship note there means unknown, never permission.
• Hiring cycles differ sharply by field. Where this field is off-cycle, the faculty count understates the year's real volume and a re-run in another season would return several times as many tenure-track lines.
• USCIS approval counts are unavailable. The Employer Data Hub returns HTTP 403 to automated access and no approval data has been published. Nothing was estimated in their place.
• No independent cross-check exists. h1bdata.info's holdings end in December 2022, so the DOL disclosure files stand alone.
• Some institutions petition through umbrella entities, research foundations and system-wide bodies, and cannot be split by campus. A campus that shows nothing may simply be filing under another name.
• Postings close without notice. Confirm every deadline on the posting itself before investing time.

Figures carrying a dashed evidence chip are drawn from our working dataset rather than the printed July edition. We publish them because they are informative, and we mark them because they have not been through the same review as the rest. Ask us before citing one in something formal.

Four places where our own evidence overrode an earlier decision of ours, including one correction to material we have already published. Verdicts 01 through 03 appear in context above. The fourth belongs here because it is a commitment to readers rather than a finding.