The offer went out in March. The candidate, a machine-learning engineer finishing a doctorate at a Midwestern university, accepted within a day. The hiring manager celebrated in Slack. The recruiter moved on to the next requisition.

In August, a short email arrived. He was grateful, he wrote, but he had decided to take a role in Toronto. Nobody at the company asked why, and he didn't volunteer it. In the applicant tracking system, the record was closed with two words: candidate withdrew.

The real reason was a visa timeline that nobody on the hiring side had ever mapped out for him, and a competitor, in another country, that could promise him a start date.

That record is a composite, but anyone who has recruited technical talent in the past five years will recognize it. It is also, in a sense, invisible. It doesn't show up as a lost hire, a failed offer or a retention problem. It shows up as nothing at all.

The attrition nobody measures

Talent teams measure almost everything. Time to fill, offer acceptance rate, quality of hire, ninety-day retention, cost per hire. Dashboards glow with them.

What almost no dashboard captures is how many strong international candidates are lost to the immigration system itself: people who accept, wait, grow anxious and quietly leave, or who never apply because the process looks too uncertain from the outside. When they go, the loss is coded as something else. A withdrawal. A counter-offer. A "cultural fit" issue. A resignation eighteen months in that the exit interview attributes to "growth opportunities."

This is silent attrition, and for companies competing for AI talent, it may be one of the most expensive leaks in the hiring funnel. It is hard to fix precisely because it is hard to see.

The clock is the problem

For most U.S. employers, the default route for a foreign-born hire is still the H-1B visa, followed years later by an employer-sponsored green card. Every stage of that route now runs on a clock the employer does not control.

The H-1B itself remains a lottery. For fiscal year 2027, USCIS replaced the purely random draw with a weighted selection process that gives more chances to higher-paid positions. That may help senior hires. It does little for the new graduate whose start date now depends on how a salary band maps onto a wage level.

Then there is policy volatility. A 2025 presidential proclamation imposed a $100,000 payment on certain new H-1B petitions. A federal court has since vacated the policy implementing it, and a new proclamation in September 2026 extended it for another year, even as the litigation continues. For a candidate weighing two offers in two countries, the details matter less than the headline: the rules might change again before they arrive.

The green card stage is slower still. The Department of Labor's own processing data shows PERM labor certifications taking around eleven months at analyst review as of August 2026, before a single immigrant petition has been filed. For Indian and Chinese nationals, per-country limits add years or even decades on top, depending on the category.

Put those numbers in front of a 29-year-old engineer with an offer from London or Toronto, and the decision starts to look less like a leap of faith and more like arithmetic.

Why AI talent is the most exposed

It would be easy to treat this as a niche compliance issue. The labor market data suggests otherwise.

The Bureau of Labor Statistics projects employment of data scientists to grow 35% between 2025 and 2035, more than ten times the 3% average for all occupations. Much of the pipeline for those roles runs through American universities, and much of it is foreign-born.

Crucially, that talent wants to stay. A National Science Foundation study of early-career science and engineering doctorate recipients who held temporary visas at graduation found that 72% were still in the United States years later. Among those from India, the stay rate was 84%; among those from China, 87%. Computer and mathematical scientists made up roughly half of the temporary visa holders the study tracked.

In other words, the U.S. does not have a problem attracting this talent. It has a problem keeping the door open long enough for employers to hire it. And the companies that fill the gap first, in Canada, the UK or elsewhere, rarely give it back.

Three patterns we see constantly

The offer that expires in the lottery. A candidate accepts in spring, is not selected in the H-1B draw and has no fallback plan from the employer. By the time HR begins exploring alternatives, a rival has already made an offer with a clear start date. The company loses not just the hire but the months spent recruiting them.

The green card holdout. An employee stays for years because leaving would reset their green card process. On paper, retention looks excellent. In practice, they are disengaged, underpromoted and privately interviewing. When a competitor offers a faster immigration route, often through a category the original employer never considered, they are gone within a quarter.

The unanswered question. A senior candidate asks in the first interview, "What's your approach to green card sponsorship?" The recruiter doesn't know. The hiring manager says it's "handled by legal." The candidate hears something different: this company hasn't thought about my future.

What leading employers do differently

The companies that handle this well tend to share one habit: they treat immigration as part of talent strategy, not a back-office compliance task.

They map a candidate's likely immigration timeline before the offer, not after. They build fallback plans for H-1B non-selection. They start green card processes early instead of rationing them as retention rewards. And critically, they know that the default route is not the only one.

Several faster lanes exist for exceptional talent. The O-1A visa is not subject to the H-1B lottery. On the permanent side, the EB-1 first preference categories include a route for outstanding researchers and an extraordinary ability category, EB-1A, that professionals can file for themselves, without a PERM labor certification or a job offer.

That last point is underused. Many senior engineers, researchers and AI specialists already have records that could support an EB-1A petition, but neither they nor their employers have ever assessed them. Specialist firms such as EB1A Experts work with professionals on exactly these profiles, and an employer that simply points its best people toward that option is offering something a competitor's H-1B promise cannot match: a credible, faster path to staying.

The real cost of silence

Every silent departure carries a bill: sourcing fees, interview hours, relocation packages, lost months of productivity and, in AI roles, delayed projects that competitors ship first. None of it lands on the immigration budget, because the immigration budget was never where the problem was recorded.

The fix does not start with legal. It starts with measurement. Add an immigration field to your exit data. Ask withdrawn candidates the question nobody asks. Track how many offers to foreign-born candidates are lost between acceptance and start date. The number will probably be larger than anyone expects, and it will be the most persuasive business case your talent team has ever brought to the C-suite.

Frequently asked questions

How do immigration delays affect AI talent retention?

They extend the gap between offer and start date and add years of uncertainty to long-term plans. Candidates with competing offers from countries with faster visa systems often choose certainty, and existing employees may leave for employers who offer quicker routes to permanent residency.

Is the H-1B still a lottery?

Yes, but since fiscal year 2027 USCIS has used a weighted selection process that gives higher-wage positions more entries. Selection is still not guaranteed, so employers need fallback plans for candidates who are not chosen.

What is the O-1A visa, and why does it matter for tech hiring?

The O-1A is a temporary visa for people with extraordinary ability in fields such as science, business and technology. It is not subject to the annual H-1B cap or lottery, which makes it a useful option for senior or highly recognized hires.

Can employees get a green card without employer sponsorship?

Some can. The EB-1A extraordinary ability category allows professionals to self-petition, with no job offer or PERM labor certification required. It suits people with a sustained record of recognized work in their field.

Should HR track immigration-related attrition separately?

Yes. Most systems code these losses as withdrawals or voluntary resignations, which hides the pattern. Adding an immigration field to offer-decline and exit data is a simple way to measure the real cost.

When should employers start the green card process for a new hire?

Earlier than most do. Because labor certification alone can take close to a year, many employers now begin planning at the offer stage rather than treating sponsorship as a reward for long tenure.

Conclusion: the leak you can actually fix

The global race for AI talent will not be won by the companies with the flashiest offers. It will be won by the ones that give exceptional people a believable answer to a simple question: can I build my life here?

Most employers already have the candidates. What they lack is a plan that starts before the offer letter and a willingness to look beyond the default visa route. The companies that get this right will stop losing talent they never knew they had lost.

If your team is hiring or retaining senior international talent and wants to understand whether faster routes like EB-1A could apply, book a consultation with EB1A Experts. It is a quick way to turn a hidden cost into a strategy.

This article is for general information only and is not legal advice. Immigration rules change frequently; employers should consult qualified immigration counsel about specific cases.