Your Hiring Plan Is Not an Immigration Strategy: What the 2026 H-1B Whiplash Means for Engineering Teams

Ten Months, One Hiring Channel, a $100,000 Swing

Put the last ten months on a single timeline and the planning problem announces itself without any commentary needed.

In September 2025, a presidential proclamation imposed a $100,000 supplemental fee on certain H-1B petitions, applying to petitions filed on or after 21 September 2025 — principally where the beneficiary was outside the United States and did not already hold H-1B status. On 29 December 2025, DHS published a final rule replacing the random cap lottery with a weighted selection process that prioritises higher-paid beneficiaries. That rule became effective on 27 February 2026, in time for the FY2027 registration window that ran 4–19 March 2026.

Then, on 8 June 2026, the U.S. District Court for the District of Massachusetts vacated the proclamation. Judge Leo T. Sorokin's reasoning was that the $100,000 payment was not an immigration restriction at all but a tax, which the president has no unilateral authority to impose, and that the agency actions implementing it breached the Administrative Procedure Act. The government moved for a stay pending appeal on 18 June; the First Circuit declined to grant it. As of late July 2026 the fee is not being collected, the appeal is live and expedited, parallel challenges are pending in other federal courts, and most practitioners expect the question to end up at the Supreme Court.

Now read that as an engineering leader rather than as a citizen. Between September 2025 and July 2026, the marginal cost of bringing one named senior engineer into a US office moved from roughly $10,000 to roughly $110,000 and back to roughly $10,000, on a schedule set by litigation nobody in your organisation controls. That is not a policy debate. That is a supply channel with six-figure variance and no forward guidance, and it sits underneath a headcount plan that your CFO expects you to commit to for four quarters.

The honest response is not to guess which way the appeal goes. It is to notice that a channel behaving this way should no longer be load-bearing in your staffing model — whichever way it resolves.

The Number That Actually Matters Is 211,600

While the fee litigation absorbed the attention, the change with real staying power arrived as a regulation — and regulations do not evaporate when a proclamation is struck down.

Under the weighted selection rule, an employer registering a beneficiary must now identify the intended geographic area of employment, the occupational category, and the corresponding OEWS wage level. Selection chances then scale with that wage level: a Level IV offer receives four entries in the lottery, Level III receives three, Level II receives two, and Level I receives one. The beneficiary-centric structure introduced in earlier cycles remains, so the change is not about fraud prevention. It is a deliberate reweighting of who gets access.

The market answered immediately. FY2027 registrations fell to 211,600 from 343,981 the previous year — a 38.5% decline, the steepest single-year drop the programme has recorded. And the composition shifted exactly as designed: only 17.7% of selected registrations sat in the lowest wage category, with a materially larger share going to beneficiaries holding US master's degrees or higher.

Translate that into a hiring funnel. The H-1B route was never primarily how companies hired principal engineers — those people usually arrive through O-1, L-1, existing status, or an acquisition. It was how companies converted the international graduate pipeline: hire on OPT, sponsor at Level I or Level II, keep the person. That conversion path is now structurally disadvantaged in a lottery where a Level I registration competes against Level IV registrations holding four tickets to its one. The pipeline did not get more expensive. For entry and early-mid roles, it got substantially narrower, and it stays narrow regardless of what the First Circuit decides about the fee.

This compounds a dynamic we have written about repeatedly: AI tooling has already suppressed junior hiring, and the people who would have become your senior engineers in 2031 are not being hired in 2026. Immigration policy has now added a second constraint on the same cohort, in the same direction, at the same time.

The Cost Stack Was Already an Argument Before Anyone Added $100,000

It is worth reconstructing what sponsorship costs when no emergency proclamation is in play, because the baseline is frequently underestimated by the people writing headcount plans.

The mandatory government fees run roughly $1,805 to $2,555 — a $555 USCIS filing fee, a $500 fraud prevention and detection fee, and an ACWIA training fee of $750 or $1,500 depending on employer size. On top of that sits the asylum program fee at $600 for employers with 26 or more full-time employees, and a $4,000 Public Law 114-113 fee for companies with more than 50 employees where over half are on H-1B or L-1 status. Legal representation typically adds $2,000 to $5,000. Premium processing, which most teams treat as non-optional when a start date matters, rose to $2,965 on 1 March 2026. The realistic all-in filing range lands somewhere between $4,805 and $12,160 per petition.

That is the visible part. The obligation that dominates the economics is the prevailing wage — the salary must be the higher of the area prevailing wage or the actual wage paid to comparable employees. And the new selection rule creates a perverse incentive around exactly this number: registering at a higher wage level improves your lottery odds, which means the rule quietly pressures employers to raise offers to buy probability. You are no longer just paying fees. You are bidding.

Then there is the calendar, which is the part that breaks roadmaps rather than budgets. You register in early March. Selection is announced weeks later. You file. Approved cap-subject employment starts 1 October. If your engineer is not selected, you wait a year — and you found out in April that a role you scoped in January will not be filled that fiscal year. Forrester's assessment of the fee's long-run effect was that it would push IT service prices up and work offshore, and the Richmond Fed made the same structural point in more neutral language: raising the marginal cost of hiring foreign specialty workers pushes employers to substitute across margins — more offshoring to foreign affiliates, more remote contracting, more automation, and a shift toward jurisdictions with predictable immigration regimes.

Note the phrase doing the real work in that sentence: predictable immigration regimes. Not cheap. Predictable. Nobody restructures a global engineering organisation to save $10,000 per head. They restructure it because they cannot forecast, and forecasting is most of what running an engineering organisation actually is.

Europe Is Running the Same Play With Different Numbers

European engineering leaders reading the H-1B coverage as American political theatre should check their own numbers before feeling comfortable, because the UK has been tightening the same channel on a parallel schedule.

The general minimum salary threshold for new Skilled Worker applications rose to £41,700 from £38,700 for applications made from 22 July 2025 — an 8% jump, applied to a threshold that had already been raised sharply the year before. The PhD-relevant threshold moved to £37,500, and the PhD STEM and Immigration Salary List threshold to £33,400. More consequentially for volume hiring, the eligibility bar was lifted to RQF Level 6 — degree level — which removed more than 180 occupations from sponsorship eligibility unless they appear on a Temporary Shortage List.

The fee stack moved with it. The Immigration Skills Charge increased 32% for medium and large organisations to £1,320 per sponsored year from 16 December 2025, and £480 for small or charitable sponsors. From 8 April 2026, a standard Skilled Worker visa costs £819 for up to three years from outside the UK or £1,618 for a longer grant, with in-country extensions higher still. Practitioners estimate the salary threshold change alone adds £18,000 to £35,000 in annual employer cost per sponsored worker before a single fee, charge or health surcharge is counted.

The pattern across both jurisdictions is the same even though the politics differ: relocation-based hiring is being repriced upward and restricted at the entry and mid levels, in both cases faster than engineering budgets are growing. A CTO in London and a VP Engineering in Austin are now solving the same problem with different paperwork.

Where the Work Went: The Capability-Centre Answer, and What It Costs to Copy

The largest companies did not wait for the litigation to resolve. They moved the work to the people instead of the people to the work — and the numbers are no longer small.

India's global capability centre ecosystem reached roughly 1,760 centres employing more than 1.9 million professionals and generating $64.6 billion in revenue by the end of 2025, growing at 11–13% annually in new centre setups, with projections of 2,100–2,200 centres and 2.5–2.8 million employees by 2030. More than 78% of newly established centres now name AI/ML and data analytics as their core capability focus — this is not the back-office captive model of 2010. CIO's reporting frames the causal link plainly: restrictive H-1B policy is driving tech talent back to India and reshaping global IT, and expanding capability centres became the obvious way for firms to keep serving multinational clients without visa exposure.

It is a genuinely effective strategy. It is also the most capital-intensive option on the menu, and this is where a lot of mid-market engineering leaders read the trend piece and quietly misfile it as advice. A captive centre means a legal entity, local employment law, payroll, benefits, real estate, an HR function, a country manager, and a twelve to eighteen month runway before the first sprint that matters. The economics work above a certain sustained headcount — typically well north of fifty engineers — and are actively punishing below it. If your engineering organisation is forty people and you need six more, a capability centre is not the answer to your problem; it is a different, larger problem you would be adopting voluntarily.

The second limitation is one the GCC boom does not solve at all: distance. The reason the offshore model generated coordination overhead was never the legal structure. It was the working hours. A capability centre with a nine-hour gap has exactly the same standup problem, the same one-round-trip-per-day decision latency, and the same tendency to drift toward specification-driven work — which, in an era where agentic tooling has made ambiguity resolution the expensive part of engineering, is precisely the wrong shape.

The Third Option: Hire Where the Engineers Already Are

Between sponsoring relocation and building a captive entity sits the option most mid-market teams underweight, largely because they last evaluated it when it was still framed as cost arbitrage: engaging a senior, dedicated team in a jurisdiction that already has the engineers, already has the legal infrastructure, and already sits in your working day.

The market has been moving this way independent of immigration policy. Nearshore software development is growing at a CAGR above 12%, more than 35% of businesses in Western and Nordic Europe plan to increase nearshore usage over the next two years — with scalability, not price, cited as the primary driver by half of them — and Eastern Europe produces over 200,000 IT graduates a year. Europe's IT outsourcing market is projected to reach roughly $282 billion by 2030.

Serbia is a useful specific case because the growth is documented rather than asserted. ICT exports reached $4.1 billion in 2024, having grown more than tenfold in a decade from under €400 million, with ICT exports at 5.02% of GDP — the most export-oriented tech sector in the Balkans, up from 2.94% in 2020. Employment in the sector grew from roughly 30,000 to nearly 120,000 professionals between 2014 and 2024 across more than 4,000 companies, and the pipeline behind it is intact: ICT graduate numbers rose almost 31% between 2020 and 2024, with around 15,000 STEM graduates entering the workforce annually.

The strategic point is not that Serbian engineers are cheaper than American ones, though they are. It is that engaging one requires no lottery, no wage-level bidding, no October start date, no proclamation, no appeal, and no exposure to a First Circuit ruling. The engineer is already legally employed, already in a jurisdiction with a stable framework, and already working in a timezone that overlaps your day. The cost is knowable in advance and stays knowable, which is the property the visa channel has lost.

There is an honest caveat worth stating rather than glossing. For a European client, Serbia is nearshore in the full sense — CET, same business day, a two-hour flight. For a US client it is not geographically nearshore, and anyone who tells you otherwise is selling. What it is, for a US team, is a stable and overlapping arrangement: Belgrade afternoons cover US East Coast mornings, which yields three to four hours of genuine same-day collaboration — considerably more than the offshore alternative, considerably less than a team in Mexico City. Choose on the basis of what your work actually requires, not on the basis of which word appears in the vendor's homepage headline.

What This Actually Changes About How You Plan Headcount

The practical work here is not immigration strategy. It is separating your open roles into categories you probably have never formally separated, and then routing each one differently.

Start by asking which roles genuinely require physical presence in a specific country. Some do: roles carrying regulated licences, roles requiring security clearance, roles whose value is in-person customer contact, roles embedded in a hardware lab. For those, sponsorship remains the tool, and you should now budget it as a probabilistic expense rather than a line item — model the lottery odds at your realistic wage level, model the year-long delay if you lose, and price the delay in delivered roadmap rather than in recruiting fees.

For everything else, the question flips. If the work is a distributed-systems backend rebuild, a payments integration, an internal platform, a data pipeline, a mobile client — the location requirement was almost always inherited from an org chart rather than derived from the work. The relevant constraints are timezone overlap, seniority, and continuity of the same people over quarters. None of the three is meaningfully improved by relocating a person, and one of them is actively harmed by a hiring process that can stall for twelve months on a lottery outcome.

Second, stop treating the collaboration infrastructure as a per-engagement cost. Teams that have already invested in written decision records, asynchronous-first communication, a documented onboarding path, and clean environment provisioning can add a distributed team in weeks. Teams that run on hallway context and tribal knowledge cannot, and will blame the model rather than the preparation. If you have been putting off that investment, immigration volatility is as good a forcing function as you will get.

Third, be sceptical of your own fully-loaded cost model. Most comparisons put a sponsored engineer's salary against an external team's rate card and stop there. The comparison is only honest if it includes filing and legal costs, premium processing, the prevailing-wage premium the weighted lottery pressures you into, the recruiting cycle, the expected value of a failed selection, and the roadmap cost of a role sitting open from April to the following October.

How We Structure This at StepTo

We are a Belgrade-based nearshore provider and have been running senior-led engineering teams from Serbia since 2014, which means our answer to this is structural rather than clever. It is worth stating plainly, including where it does not apply.

Our engineers are our direct employees, working from Serbia, on a single legal entity, with no visa dependency in the arrangement at all. When a client engages a dedicated development team, they get named people assigned to their product — not a rotating bench, and not a subcontracting chain assembled after the contract is signed. That matters more in 2026 than it did in 2020, because the same property that removes immigration exposure also happens to be what European regulated clients now need to declare in their supplier registers.

The predictability point is the one we would emphasise to anyone re-planning headcount this quarter. A team engaged this way has a cost you can put in a budget and a start date you can put in a roadmap. There is no March registration window, no wage-level bidding to buy lottery probability, no April disappointment, no 1 October dependency, and no exposure to what an appellate court decides about the taxing power of a proclamation. Whatever the First Circuit does with the $100,000 fee, it does not change your delivery date.

On timezone, we say the same thing to prospects that we said above: Serbia is genuinely nearshore for European clients and time-shifted-but-overlapping for US ones. We are candid about that in the first call because the alternative — discovering in month two that a client needed six hours of overlap and has three — is a bad outcome for both sides. For work where continuity, seniority and same-day European collaboration matter, this model is hard to beat. For a US team that needs full-day pairing with a San Francisco product org, Latin America is a better geographic answer and we will say so.

And for teams weighing this against the captive-centre route: engaging an established team in Serbia gets you into the same talent market without the entity, the payroll infrastructure, the local HR function or the eighteen-month runway. If the engagement later grows into something worth owning outright, a build-operate-transfer path exists. Starting with the entity, before you know whether the model works for your product, is the expensive way to find out.