The Student-to-Founder Pipeline Built America's Unicorns. In 2026 It's Narrowing.
I run our university and student programs, which means I spend most of my week on two phone calls that sound like they're about completely different things.
The first is with a twenty-year-old who has already built something. Not an idea — a product, a waitlist, sometimes revenue. They want to know how to get in front of a real founder, a real investor, a real market. The second is with a university or an innovation agency asking, more or less, the same question in institutional language: our students are more entrepreneurial than any cohort we've had; how do we stop losing them?
Both calls are symptoms of the same thing. The pipeline that turns students into founders is under more pressure in 2026 than at any point in the last decade — and more people are trying to get into it than ever. Here's the data on both halves of that sentence, and what I think we should actually do about it.
The generation that stopped waiting
Figure 1 — Youth entrepreneurship at a record: U.S. 18–24 early-stage activity is level with 25–34 for the first time.
Start with the demand side, because it's the part that's genuinely new.
For most of the last twenty years, entrepreneurship in the U.S. was a mid-career act. You learned in someone else's company, then you left. That has inverted. In the Global Entrepreneurship Monitor's latest U.S. data, 18-to-24-year-olds now show a total early-stage entrepreneurial activity rate of 25% — tied with 25-to-34-year-olds for the first time since GEM began measuring it. In 2013 that youth figure was around 10%. It has roughly two-and-a-half-x'd in a decade.
Look inside the cohort and it gets sharper: 24% of 18-to-24-year-olds are already running something, and another 21% intend to start within three years. And this isn't an American quirk. GEM finds young people leading the rate of business creation in 42 of 51 economies studied — 80% of the world.
I want to be careful about what that number is. A 25% early-stage rate is not 25% future unicorn founders; it counts a lot of side projects, a lot of solo consultancies, and a lot of things that will quietly close. But that's exactly why it matters to anyone building programs. It means the raw willingness is no longer the constraint. We are past the era of persuading students to try. The constraint has moved downstream — to whether any of it connects to a market.
What that pipeline has already built
Figure 2 — Almost one in four U.S. unicorns has a founder who arrived as an international student.
Now the supply side, and the number that should end the argument about whether this matters.
The National Foundation for American Policy published its 2026 count of U.S. billion-dollar companies in June. Of the 775 U.S. unicorns as of April 2026, 455 — 59% — have at least one immigrant founder. Their collective value has gone from $168 billion in 2016 to $5.0 trillion in 2026.
Then the line I'd put on a wall in every international office in higher education: almost one in four of America's unicorns — 24%, or 183 companies — has a founder who first came to the United States as an international student. Those 183 companies are worth somewhere between $3.5 and $4 trillion, and each one has created an average of 1,123 jobs. The founders come from 76 different countries; India alone accounts for 96 of the companies.
Read that mechanically rather than sentimentally. A student visa was the entry point for a quarter of the most valuable private companies America has. Not the alumni network, not the venture capital, not the accelerator — those all came later. The entry point was a classroom. And the same body of research finds roughly three-quarters of the foreign-born founders of venture-backed U.S. startups arrived as students.
Which is why what happened to enrollment this year is a startup story, not just an education story.
The door is narrowing
Figure 3 — The total looks calm; the intake does not. New enrollments down 17%, graduate down 12%.
The headline number looks calm. The one underneath it doesn't.
IIE's Open Doors 2025 recorded 1.2 million international students in the U.S. in 2024/25 — an all-time high, up 5%. But new enrollments, students arriving for the first time, fell 7% to 277,118. A record total built on a shrinking intake.
The Fall 2025 Snapshot, covering the 2025/26 year across more than 825 institutions, shows what happened next. Total international enrollment is down only 1%. But:
New enrollments are down 17%.
Graduate enrollment is down 12% — the exact population that staffs the labs and produces the deep-tech spinouts.
Undergraduate numbers are up 2%, and OPT is up 14%, but that OPT growth is the tail of previous strong graduate years working itself through the system. It's a trailing indicator, not a healthy one.
57% of institutions report a decline in new international enrollment. Among them, 96% cite visa application concerns and 68% cite travel restrictions.
72% are offering deferrals to spring 2026, and 56% to fall 2026 — institutions holding places for students who couldn't get in the door on time.
The policy environment around that is genuinely tighter, and I'll state it plainly and without editorializing, because founders and program directors need to plan against facts. A $100,000 fee on new H-1B petitions, imposed by proclamation in September 2025, survived a court challenge in December 2025 and is scheduled to run through at least September 2026 unless extended. A wage-weighted H-1B selection system took effect on February 27, 2026, improving odds for higher-wage registrations — which structurally disadvantages entry-level roles, meaning recent graduates. For F-1 and J-1 students, "duration of status" has been replaced with a fixed admission period tied to the I-20, and the post-completion OPT grace period has been cut from 60 days to 30. The I-765 filing fee rose to $1,780. From January 1, 2026, F, M and J visa issuance is partially suspended for nationals of 19 countries under Proclamation 10998. A $100,000 OPT fee has been floated but not enacted — worth watching, not yet worth planning around.
Institutions have not lost interest: 84% still call international recruitment a priority, and 78% report equal or higher funding for it. And 92% say that without OPT, international students would simply choose other countries. That last figure is the whole risk in one line. The pipeline doesn't disappear when a door narrows. It reroutes.
Seven thousand accelerators, and the one variable that predicts results
So the instinct is to build a program. Fine — but let's be honest about the field we're building into.
There are now more than 7,000 accelerator programs worldwide, a market worth $5.11 billion in 2025 heading to $6.07 billion in 2026, growing at roughly 8.8% a year. North America holds about 41.6% of it; Asia-Pacific, at 27.7%, is growing fastest. Incubators, campus pre-accelerators, government soft-landing schemes, corporate verticals — the supply of programming is not the problem either.
The quality of it is. A meta-analysis published in the Journal of Technology Transfer this year pooled 21 studies and 68 effect sizes and found a statistically significant positive effect from accelerator participation — real, but conditional. The effect is moderated by program duration, cohort size, sponsorship type and regional context. And the finding I care most about: effectiveness varies with the density of the surrounding ecosystem and with the degree of university involvement.
Put those two facts side by side. Seven thousand programs; a measurable effect that depends heavily on whether a university is genuinely wired in. Most programs are not. They run adjacent to campuses, recruit from them, and share nothing structural with them.
And then there's the failure mode I wrote about earlier this year looking at Latin America, which applies to every region I work in: most acceleration still stops at the border. It gets a company to local product-market fit and waves goodbye as the founder boards a flight to raise somewhere else. That isn't a bridge. It's a boarding gate.
My colleague Anastasia has made the institutional version of this argument: a campus can manufacture talent, IP, density and capital and still leave its startups trapped inside one national market. And as we've written about the corporate innovation stack, programs only produce outcomes when they're wired to someone who can actually buy, pilot or hire.
What we actually built
Figure 4 — A bridge, not a boarding gate: the Elpis Labs internship join, and one named outcome.
This is the part of my job I can describe concretely rather than theoretically, so here it is without the marketing gloss.
The Elpis Labs Internship Program exists because we needed the two halves of our own business to touch. On one side we run U.S. market-entry and acceleration programs for international startups and public agencies — Korean deep-tech cohorts in New York through the KSC Deeptech program, the Uzbek Startup Summit, the Korea–NYC Venture Showcase, an MOU with Hanyang University's Startup Support Group signed at CES 2026. On the other, we work with universities, accelerators and innovation hubs across the U.S., Europe, MENA and Central Asia. The internship program is the join.
In 2025 we matched more than 50 students. They came from our associate universities — Princeton, Columbia and Columbia Business School, NYU Stern, Northeastern, Boston College, Rutgers, Fordham, Baruch, Hunter, the City College of New York and Montclair State — which is to say: a mix of Ivies, a business school, two big state universities and four CUNY and commuter schools. That mix is deliberate. The single most over-served population in startup programming is the student who was already going to find their way in.
It runs on two tracks:
Accelerator Associate — you work inside our accelerator and venture-studio teams: product discovery, market research, pitch preparation, go-to-market strategy, U.S. market entry. Directly with early-stage founders building globally scalable products.
Venture Capital Associate — you work with our partners on startup sourcing, deal analysis, ecosystem research and investment theses across AI, Web3, deep tech and emerging markets.
The mechanics matter more than the mission statement. Eight to twelve weeks. Ten to twenty hours a week. Hybrid and remote-first, with on-site participation at events and partner programs in NYC, Europe, MENA and Asia. After application and interview, interns are matched to a specific startup or internal team based on skills and interests — matched, not pooled. Requirements are a strong motivation, an analytical mind and fluent English; a background in business, tech, economics or design helps but isn't mandatory. Supervision comes from our Managing Partners, Anastasia Lykova-Allan and Kenneth Huynh, and our Program Director, Gabriel Arant.
The university side isn't only internships, either. We bring student and researcher cohorts into rooms their campuses can't book on their own. A cohort from KAIST and Chung-Ang University came through New York and spent time inside Mastercard's Experience Center, working through the experiential demos of how a global payments network is actually deploying AI across its ecosystem — not a lecture about corporate innovation, the thing itself, with the people who run it. Our own cohorts have sat inside Bloomberg and walked the hardware ecosystem at New Lab in the Brooklyn Navy Yard.
That is the specific scarcity worth naming. A strong university can manufacture talent, research, density and even capital. What it cannot manufacture is an afternoon inside a Fortune 500's innovation function. Corporate access is the one input that has to be brokered, and it's the input that turns an interesting student project into something a company can pilot.
We also publish the work rather than gatekeeping it. Our video library is open: sector-by-sector breakdowns of corporate venture capital in banking, payments, automotive, aviation, health and wellness; company deep-dives on JetBlue Ventures and BMW i Ventures, Deutsche Bank Ventures, Natura, Zara, Cargill, Mondelez; and program films from the Italian Global Startup Program and the Miami Softlanding Program. For a student trying to understand how a corporate actually evaluates a startup, that's a curriculum nobody assigns you.
The job the students actually do
Figure 5 — Four students, four universities, one job description: the student as translation layer.
Here's the thing I didn't expect when we started, and it's the most useful thing in this article. I went back through what our interns say about their placements, and almost all of them describe the same job — in their own words, unprompted, across different universities and completely unrelated sectors.
Chrystee Xing, a student at NYU, was our Global Innovation & Product Strategy intern: "I learned how to adapt global products for local markets and developed a sophisticated understanding of how startups scale." She got there by being "connected with international founders who are bringing world-class technology to the US."
Pratyus Mohapatra, at Princeton, worked with a South Korean unicorn: "[it] allowed me to bridge the gap between their advanced technical data and the US market, specifically by helping them package and sell their product through better communication."
Muhammad Salman, at Montclair State, describes it almost identically: "helping a high-growth startup translate their complex data capabilities into a sellable package for the US market."
And Francis LiButti, at Rutgers – New Brunswick, was matched to SG Labs, a Korean sports-tech startup, where he spent the summer on marketing automation and applied AI tooling to scale the company's reach in a market it was entering for the first time:
"My internship with SG Labs, a Korean sports-tech startup, was the defining moment of my transition from student to professional… The program's networking power is incredible — you aren't just interning for one company; you are part of a cohort of innovators. This experience was so successful that I am officially joining the team full-time in their New York City office this fall. Elpis Labs doesn't just offer internships; they offer career pathways."
Adapt. Bridge. Translate. Package. Four students, four universities, and one job description nobody wrote down.
The student is the translation layer. That's the finding. A foreign company arriving in the United States has a technology problem it has already solved and a legibility problem it hasn't: its data is impressive and unreadable, its product is excellent and mispositioned, its pitch is accurate and unpersuasive to an American buyer. What it needs is somebody who is genuinely native to the destination market and genuinely comfortable with the technology — which is a fairly precise description of a good student in New York.
So the value doesn't flow one way. We used to talk about this program as giving students access. It is at least as true that the students are giving the companies something they cannot buy: a reader. And it explains the outcome — Francis was hired by the startup he was assigned to, because by week ten he wasn't an intern, he was the person who knew how to make the company make sense here.
It also isn't only Korea and it isn't only software. Shiema Deffallah, at Northeastern, worked with a clean-tech company on mobile battery solutions and came out with something more elementary and probably more durable: "Even early in my academic career, Elpis Labs pushed me to learn the 'language of business'… The program forced me out of my comfort zone, teaching me how to network effectively with investors and stakeholders from diverse backgrounds."
One honest note, because I'd rather you hear it from me than discover it: the remote-first element is the part interns find hardest. Salman called the remote collaboration "a valuable learning curve" and singled out the in-person training on networking and pitching as the highlight. That tracks with everything else here. The translation work needs proximity, and the parts of our program that happen in a room are the parts that compound.
The playbook
If you run a university or an innovation agency: your students' entrepreneurial intent is at a record and your new international intake is under pressure. Those two facts point at the same intervention. Stop treating the outbound student and the inbound startup as separate portfolios. Place your students inside foreign companies entering your market, and your local companies inside cohorts abroad. The meta-analysis says university involvement is what makes acceleration work — so be involved structurally, not adjacently. And build the bridge before you need it: the pipeline reroutes long before the statistics show it.
If you're a student: your degree buys you talent, credentials and density. It does not buy you a market. Go get proximity to one — to founders making decisions, to investors saying no and explaining why, to a company entering a country for the first time. And understand what you're actually worth to them: not cheap labor, but a reader. You know how this market hears things, which is the one thing a foreign founder cannot acquire on a plane. Twelve weeks inside a real go-to-market beats a year of case competitions. Know the visa mechanics that apply to you and plan two years out, not two months. Pick programs by what you'll touch, not by the brand on the certificate.
If you're a corporate or an investor: the cheapest R&D you will ever run is a cohort of students working on a market-entry problem you haven't staffed. And the cheapest recruiting channel you will ever run is the same cohort. Francis was hired by the startup he was assigned to. That happens more often than the sector admits, because a twelve-week matched placement is a better hiring signal than any interview loop.
How Elpis reads it
The mistake is thinking of the student-to-founder pipeline as a destination — a country you get into, a campus you get onto, a visa you hold. Framed that way, 2026 is unambiguously bad news, and there's nothing to do but wait for policy.
We think it's a network. And networks respond to narrowing by rerouting, which means the work is to build more edges: cross-border internships, cohort placements, university partnerships that run in both directions, programs that follow the founder into the market instead of waving from the gate. 24% of America's unicorns started with a student arriving from somewhere else. The interesting question for the rest of this decade isn't whether that talent still exists. GEM settled that — it has never been more willing.
The question is who is building the connections that let it reach a market at all. And the fifty-odd students we matched last year suggest something more useful than charity: the student isn't only the person the pipeline is for. Increasingly, the student is the pipeline.
So: which universities are actually wired into your ecosystem — and does the wiring go both ways? Tell me in the comments. And if you're a student who wants to be in the game rather than watching it, our program is open.
Comments