H-1B Wage Levels Now Decide Your Startup's Lottery Odds

The H-1B lottery is no longer random. Starting with the FY 2027 cap, USCIS runs a weighted selection that gives each registration a number of entries equal to its DOL wage level: Level IV gets 4 entries, Level III gets 3, Level II gets 2, and Level I gets 1. Your offered salary, benchmarked against the local government wage bands for that role, now sets how many times your candidate’s name goes in the hat. For a startup making a market-rate offer to an early-career engineer, that turns a coin-flip into a number you can model, and mostly a number working against you.

This is a compensation-benchmarking problem wearing an immigration costume. None of what follows is legal advice; confirm any specific wage level and strategy with an immigration attorney and the Department of Labor’s FLC Data Center. But the decision that actually moves your odds happens weeks before any lawyer files anything, at the moment you set the salary band. That is the decision this guide is about.

The lottery isn’t luck anymore, it’s your salary

The weighted-selection rule is final. It published in the Federal Register on December 29, 2025, took effect on February 27, 2026, and first applies to the FY 2027 cap season. Instead of one entry per registration drawn at random, USCIS now enters each registration a number of times equal to the wage level attached to it.

Here is what that does to the odds. DHS published its own projected selection rates in the rulemaking:

A Level I registration now has roughly half the odds of the old flat lottery. A Level IV has more than double. Those are DHS projections, not guarantees; the real odds in any given year depend on how registrations distribute across the four levels. But the direction is not in doubt. As immigration analyst Stuart Anderson reported in Forbes, the rule effectively re-sorts the winners by how much they are paid. Level is destiny.

One thing the weighting does not change: each unique beneficiary is still counted once toward the cap no matter how many entries the level buys. You cannot file five registrations for the same person to multiply chances. The only lever is the level, and the level comes from the wage.

This article is the benchmarking companion to our guide on H-1B alternatives after the $100K fee, which covers the O-1 visa, founder self-sponsorship, EOR routes, and the all-in cost model. If you are weighing whether to sponsor at all, start there. If you have decided to enter the lottery and want to understand the number that sets your odds, keep reading.

Where wage levels come from: the four percentiles nobody explains

A “wage level” is not a salary. It is a fixed percentile of the local government wage distribution for your specific occupation and metro area. The four levels are the same everywhere:

• Level I = 17th percentile
• Level II = 34th percentile
• Level III = 50th percentile (the local median)
• Level IV = 67th percentile

Those percentiles come from the Bureau of Labor Statistics OEWS survey (Occupational Employment and Wage Statistics), broken out by SOC code (the government’s occupation classification) and by metropolitan area. The percentiles are identical for a software developer, a nurse, and a mechanical engineer. What changes is the dollar figure each percentile resolves to, because each occupation and each metro has its own local wage distribution.

So “Level II for a backend engineer” is not a number you can memorize. It is the 34th percentile of what backend engineers earn in that metro, and it is a different dollar amount in San Francisco than in Austin. The Institute for Progress has a clear breakdown of how the percentile system maps onto real wages.

To find your level, you take the proffered wage (the gross base salary you are offering) and find the highest of the four levels that your salary equals or exceeds for that SOC code in that area. Offer above the 67th percentile and you register Level IV. Land between the 34th and 50th and you register Level II. The rule text requires that the level you register match the level on the eventual petition, so the number has to be real.

The same $150K is a different level in a different city

Because the levels are local percentiles, an identical salary can register at different levels depending on where the candidate works. This is the single most important thing for a startup to understand, and the easiest to get wrong.

Take two identical $150,000 senior-backend offers. In San Francisco, where the local wage distribution runs high, $150K might clear only Level II (2 entries, roughly 30.58% projected). In Austin, where the local distribution is lower, that same $150K can clear Level III or even Level IV (3 to 4 entries, roughly 45.87% to 61.16% projected). Nothing about the candidate, the role, or the money changed. Only the metro’s local OEWS band moved, and the odds nearly doubled.

For a concrete anchor, aggregated OFLC data via h1bgrader puts the entry-level (Level I) prevailing wage for a Software Developer around $135,699 in San Francisco and around $103,209 in New York for a recent period. Treat those as directional; the exact figure for your SOC and metro refreshes each July and lives at the FLC Data Center. The lesson is structural: benchmarking against the local band for the exact worksite is the whole game.

For a distributed startup, this cuts both ways. A remote hire assigned to a lower-wage metro may clear a higher level on the same budget. But you cannot pick a metro for lottery advantage that does not match where the person actually works; the worksite drives the wage determination, and a mismatch is exactly the kind of thing that gets scrutinized.

How your offer’s structure silently sets your level

Even when founders benchmark the right metro, three mechanical details quietly knock the level down a peg. Each one is avoidable if you know it exists before you write the offer.

Equity does not count. The proffered wage is gross base salary only. Bonuses, RSUs, and option grants are invisible to the wage level. A seed startup offering a new-grad engineer $120K base plus an option grant it values at another $60K feels like a $180K package. To the lottery it is a $120K proffered wage, likely Level I, one entry. The candidate the founder was most confident about is, statistically, the least likely hire on the team to clear the cap.

A salary range registers at the bottom. If you post the offer as a range, the lowest number generally controls the level. “Software Engineer, $135K to $175K” that you assumed clears Level III on the strength of the $175K top can register at Level I on the $135K floor.

Multiple worksites register at the lowest applicable level. If the candidate will split time across two offices, you must use the lowest applicable level across those worksites. A hybrid role spanning a high-wage and a low-wage metro takes the lower one.

Stack these and a “$135K to $175K, hybrid across SF and a second office” posting that felt competitive can silently land at Level I, a single entry at roughly 15.29% projected odds. The practitioner guidance at Manifest Law walks through each of these gotchas. The fix is not clever; it is operational. Benchmark the exact SOC and primary worksite, set a defensible single number, and register the level you actually intend.

“Just pay more” works, until it’s a red flag

The obvious move is to bump the salary until it clears the next level. Sometimes that is exactly right. Paying a genuinely higher, defensible wage into the next band does buy real entries, and where the role supports it, that trade can be worth it.

What you cannot do is inflate a junior role to a senior level to farm entries. USCIS has signaled it will scrutinize whether the wage level, job classification, and job description align, and a mismatch between the registration and the eventual petition can mean denial, revocation, or an enforcement referral. Registering a new-grad position at Level IV because you liked the odds, then filing a petition that describes an entry-level job, is the fact pattern that gets caught. Immigration counsel at Envoy Global and Reddy Neumann Brown both frame the honest version the same way: benchmark accurately, then decide, for roles that genuinely support it, whether a real raise into the next band is worth the odds it buys.

The framing that keeps you safe is accurate benchmarking, not gaming. Put the role and the band where they honestly sit, understand the odds that level buys, and change the offer only when the seniority of the actual job justifies the higher number.

The threshold could move under you

There is a live risk on top of all this, and it is important to describe it accurately: it is proposed, not law. On March 27, 2026, the Department of Labor published a proposed rule (an NPRM) that would raise the percentile floor for every level:

If that proposal is finalized, the dollar threshold for each level jumps sharply. A salary that clears Level II today could fall to Level I under the new floors, cutting your entries even though you never touched the offer. The comment period ran roughly 60 days, with comments due around May 26, 2026, and no final rule had published as of this writing. Boundless tracks the proposal, and the primary text is Federal Register 2026-06017.

Model it as a scenario, not a requirement. When you set a band today, it is worth asking a second question: if these floors take effect, does this offer drop a level? That is a planning input, not a rule to comply with, and it is one your immigration counsel will want to weigh in on.

Why this hits startups hardest

The wage-weighted lottery did not just change the math. It re-sorted the winners by employer size, and startups landed on the wrong side.

DHS’s own analysis notes that roughly 90% of H-1B applications for international students sit at Level I or II, because these candidates have limited experience. That is precisely the population a seed-stage startup hires: recent F-1 OPT grads and early-career engineers, with the “real” upside loaded into equity that does not count toward the wage level. Big Tech, filing senior roles with cash-heavy comp bands, disproportionately registers Level III and IV, and now wins a structurally larger share of the cap. The candidate profile startups love has become the one least likely to clear the lottery.

That is the uncomfortable takeaway: a market-rate junior offer from a great startup is now, by design, a Level I or II registration fighting for a shrinking pool. You cannot out-spend Big Tech into Level IV on a new-grad hire, and you should not try. What you can do is see the level clearly before you commit, so a sponsorship-dependent hire is a decision you priced, not a lottery result you were surprised by.

Benchmark the level before you make the offer

Every expensive failure in this article, the range that registered at its floor, the equity that did not count, the metro benchmarked wrong, has the same root cause: the wage level was discovered at registration, weeks after the band was already set and the candidate already courted. By then the only options are a scramble to change the offer or a coin-flip you did not choose.

The fix is to move the wage level upstream, into the comp-benchmarking step where the band is chosen. The two systems already speak the same language. Kit’s compensation research turns scraped market data into percentile bands (P25 / P50 / P75) for a role and location. The DOL wage levels are also fixed percentiles (17 / 34 / 50 / 67) of the local wage distribution. They are the same coordinate system, which means the market band you are already looking at can be annotated with its H-1B wage level directly.

That lets a founder see, at the moment the band is set, not just “market P50 for this role is $X” but “your planned offer of $X registers at Level II, 2 entries, roughly 30.58% projected, and reaching Level III would cost about $Y more.” Because the proposed DOL floors just re-map the same percentiles, the same view can show a labeled “if the proposed floors take effect” scenario next to the current one, so you can see whether today’s Level II band would slip to Level I under the proposal. Kit surfaces the numbers and the scenario; you and your immigration counsel decide. It never registers a level or files a petition for you.

The through-line is simple. The H-1B wage level is now a compensation-benchmarking output, not an immigration afterthought. It belongs on the comp model you build before the offer, right next to base, equity, and the market percentile, because it is made of the same percentiles they are.

If you want the rest of the picture, our companion guide on H-1B alternatives and the all-in cost model covers the $100K fee and the O-1 and EOR routes, our founding engineer salary guide shows what Level I through IV actually look like in real engineering comp, and our take on honest salary ranges explains why the range you post is the range that counts.

FAQ

How are H-1B wage levels calculated? Each level is a fixed percentile of the local government (OEWS) wage distribution for your occupation’s SOC code and metro area: Level I = 17th, Level II = 34th, Level III = 50th, Level IV = 67th. Your level is the highest one your gross base salary equals or exceeds. Because the dollar value of each percentile changes by occupation and city, the same salary can be a different level in a different metro.

Does paying more improve my H-1B odds? Yes, mechanically: a higher wage that clears the next level’s percentile adds an entry, and DHS projects odds of roughly 15.29% at Level I rising to 61.16% at Level IV. But you cannot inflate a junior role to a senior level to farm entries. USCIS scrutinizes whether the wage level, job classification, and description align, and a mismatch can trigger denial or revocation. Raise the number only when the actual seniority of the role supports it.

Does equity count toward the wage level? No. The proffered wage is gross base salary only. Bonuses, RSUs, and stock options are invisible to the wage-level calculation, which is why equity-heavy startup offers to early-career engineers often register at Level I even when the founder thinks of them as much larger packages.

Is the higher wage-level requirement law yet? No. The DOL rule that would raise the percentile floors (Level I to 34th, II to 52nd, III to 70th, IV to 88th) was a proposed rule published March 27, 2026, and was in its comment period as of this writing. No final rule had published. Treat it as a scenario to plan for, not a current requirement, and confirm the current thresholds with the FLC Data Center and your immigration attorney.