bonniej/iStock/Getty Images Plus
America longs to be “AI ready,” but there’s a gap between what we say we’re building and what we’re actually building. We invest billions in chips, stand up research initiatives and hold hearings about not falling behind China.
Yet, within a few weeks this summer, Washington took aim at the only two sources of advanced technical talent this country has. On July 1, new federal loan caps made it harder for working- and middle-class Americans to finance the graduate education AI work requires. And on July 17, the Department of Homeland Security finalized a rule putting a four-year timer on international students, who fill the majority of our AI-related graduate programs.
As a professor who studies AI’s impact on education and co-leads the University of Southern California’s Center for Generative AI and Society, I'll be blunt: There is no third pipeline. Advanced talent either grows here or comes here. We’re now constricting both.
The First Burning End: Pricing Out Domestic Talent
The One Big Beautiful Bill Act triggered the most sweeping changes to federal student loans in decades. New graduate borrowers after July 1 face a restrictive framework: Grad PLUS, which once covered the full cost of attendance, is eliminated. General graduate borrowing is now capped at $20,500 a year with a $100,000 lifetime limit. Professional tracks like medicine and law get a $50,000 annual ceiling, but only 29 government-approved fields qualify for that upper tier, with litigation on eligible programs ongoing.
Students building the next generation of AI systems typically need technical degrees—data science, computational social science, human-computer interaction. None qualify for the higher caps; they’re stuck with the $100,000 lifetime limit, often not enough. The result is predictable: Students who can self-fund will enroll, and the talented students who can’t, won’t. We don’t have enough of the former, and we’re wasting the latter.
For those students admitted directly into Ph.D. programs with funding, these loan caps may not make much, if any, difference, but for everyone else—including students who start on a master’s track—it could lock them out. It is also worth noting that Ph.D. cohorts have shrunk during these fiscally volatile times, so they do not serve as a sufficient buffer to the loss of talent.
The Second Burning End: Showing the Door to International Talent
The DHS rule ends “duration of status,” a policy since 1978 letting international students stay as long as their program required. Starting September 15, their stay is capped at four years unless they secure a discretionary extension, with new limits on changing majors or institutions too.
Four years sounds reasonable until you remember what advanced training takes. Almost no Ph.D. program finishes in four years. At the USC Rossier School of Education, we guarantee four years of funding but often extend to five. My own Ph.D. was funded for seven years; I finished in five, uncommonly fast. STEM Ph.Ds. can take even longer, and Optional Practical Training—the work authorization that allows graduates to apply their training at American companies—routinely pushes a student’s stay past four years.
Every one of those students must now petition for a discretionary extension from an immigration system that already carries a backlog of more than 11 million cases and average processing times of over a year, per the American Immigration Council. A fifth-year Ph.D. student doesn’t abandon her research gracefully when an extension is denied—she takes it to Toronto, London or Shenzhen.
Why Both Cuts Land on the Same Artery
Roughly two-thirds of graduate students in AI-related fields at U.S. universities are international, and many who earn doctorates here want to stay and build here. A National Foundation for American Policy analysis found that more than half of America’s billion-dollar startups had an immigrant founder, and about a quarter had a founder who arrived as an international student.
America’s competitive advantage lies in the combination of homegrown talent and the fact that many of the world’s best minds choose to come here. China graduates more STEM Ph.Ds. than we do. What China cannot outpace is a system that both develops its own people and attracts everyone else’s—or rather, it couldn’t, until we started dismantling that system ourselves.
You can’t get AI right if you don’t get the people right. Higher education is where talent pipelines converge. It’s how Americans level up and how the world’s talent becomes America’s talent. Squeeze it from both directions and the whole system contracts: Fewer domestic students can afford to enroll, fewer international students can risk it, and programs shrink or close. A closed program trains no one.
A nation that prices its own citizens out of advanced training while showing the door to everyone else isn’t practicing fiscal discipline or border security. It is unilaterally disarming in the global race for technological supremacy—from both ends at once.
