Silicon Valley has always run on a mix of sand and serendipity: sand for the chips, serendipity for the people who show up with improbable résumés and oversized ambition. The valley’s most durable fuel, though, is a steady flow of human capital from abroad—engineers, data scientists and product thinkers who arrive with degrees earned in Delhi, Shenzhen, Kraków or São Paulo and help turn promising ideas into companies, and companies into entire categories. For three decades, the H-1B visa has been the intake valve for that talent. It is easy to caricature as a corporate hiring tool; it is better understood as industrial policy by accident.

Start with the scale. Congress set the cap for new H-1B visas at 65,000 a year, with an extra 20,000 for those holding U.S. advanced degrees. The numbers haven’t meaningfully budged while the economy they serve has transformed several times over. Year after year, U.S. immigration authorities hit the cap quickly, and then the lottery shuts the door until the next cycle, leaving start-ups and giants alike to plan around a scarcity they can’t control.

The cap looks even smaller set against who actually powers the Valley. Foreign-born workers now account for roughly two-thirds of Silicon Valley’s tech workforce, a share that has climbed steadily over the past generation. In the day-to-day of software sprints and tape-outs, that means product teams are routinely stitched together with people who started life on different continents but now write in the common language of code and clock cycles.

H-1B is not the whole story; it is the hinge. Many future H-1B professionals first come as international students, join a company through the Optional Practical Training program, and—if the firm is lucky in the draw—convert to H-1B to stay. That OPT bridge extends to three years for STEM graduates, long enough to ship real products and to stick the landing on a visa petition. Without that on-ramp, a meaningful share of America’s university-trained talent would fly home just as it becomes commercially valuable.

What happens after they land is visible in patent filings and cap tables. Immigrants account for a disproportionate slice of the country’s inventive output—16% of inventors, 23% of patents, and roughly a third of overall innovation when spillovers to collaborators are included. At the company level, more than half of U.S. billion-dollar start-ups were founded by immigrants or include immigrant co-founders; by some counts, immigrant founders make up 44% of unicorn founders, with India the single biggest source country. The H-1B program is not the only path those founders took—but it is a major reason many of their earliest engineers and product leaders were in the room when it mattered.

The industrial imprint is everywhere you look. In semiconductors, the leap from gaming GPUs to the backbone of modern artificial intelligence required not just one visionary founder but legions of physical-design specialists, firmware engineers and applied mathematicians able to traverse global supply chains and niche toolchains. In enterprise software, the long migration to cloud architectures depended on database theorists from Prague, cybersecurity researchers from Tel Aviv, SREs from Bangalore and Kharkiv. Search and digital advertising matured through the work of multinational research labs that seeded ad-quality teams and ranking algorithms with graduates of IITs, Tsinghua and ETH Zürich. The particular alchemy of Silicon Valley is that these people met in the same place, at the same time, with the freedom to quit on Friday and start on Monday somewhere else—portability that has improved in recent rulemakings even as the broader system remains creaky.

Critics say the program suppresses wages or outsources core work to third-party staffing firms, and there is truth to the unevenness. Large platform companies and IT services providers dominate petitions in any given year, and new arrivals cluster in computer occupations. Yet the notion that H-1B is primarily about discount labor is hard to reconcile with the pay stubs. Median salaries for new computer-related H-1B roles eclipse six figures and scale quickly with experience. The more basic rebuttal is economic: if these skills were abundant at home, employers wouldn’t gamble product road maps on a lottery.

Country concentration fuels another worry. In recent years, a large majority of H-1B approvals have gone to workers born in India, with China second—an accurate reflection of the world’s current pipeline for advanced technical graduates but a concentration that can become a policy flashpoint, particularly in moments of geopolitical tension. The right response is not to shrink the program but to professionalize and modernize it—tighten compliance, reward genuine skill, and widen the aperture to talent from more places—because the underlying demand is not going away.

Policy volatility is the wild card. The government has begun to update program rules, shifting to a beneficiary-centric selection to curb gaming and clarifying portability and site-of-work requirements—useful housekeeping for a 21st-century labor market. Yet the same system can lurch in the opposite direction. A newly proposed $100,000 application fee per H-1B, championed by the White House, would function like a tax on R&D hiring, punishing the very start-ups policymakers say they want to incubate. Big Tech could afford it; seed-stage companies could not. The fee would not touch current holders, but it would chill the next cohort—and the next product wave.

The economic mechanism at work is simple and powerful. A single engineer rarely makes an industry. A critical mass of specialized people, colocated, can. H-1B didn’t design the chip architectures that power modern AI or write the standards behind cloud computing. It created the conditions for those things to happen in America by making it possible for the requisite human capital to gather and recombine here. In microeconomic terms, the visa acts as a catalyst for agglomeration—thickening local labor markets with skills that are scarce and complementary, raising the returns to being in the Valley as an engineer, a supplier, an investor or a customer.

There are better ways to run it. A skills-weighted selection beats a pure lottery; predictable timelines beat annual roulette; faster routes to permanent residency would anchor talent rather than forcing serial renewals. The master’s cap, which privileges U.S. degrees, should sit alongside pathways that recognize global excellence earned elsewhere. And if the country is serious about onshoring critical technologies—from advanced packaging to biomanufacturing—it should treat talent the way it treats fabs and foundries: as infrastructure worth subsidizing, not taxing.

For all the drama about offshoring, the more common reaction to visa rejection inside a U.S. start-up is not to ship the job to another country; it is to abandon the project, or slow it to the pace of hiring that can be done locally. The opportunity cost doesn’t show up in trade statistics. It shows up in the absence of the next product cycle, the missed standard-setting window, the European or Asian competitor that wins the category. The flip side is just as invisible: every time a graduate on OPT clears the H-1B hurdle, a Bay Area team locks in a capability it could not otherwise find, and an industry you will recognize three years from now moves a little closer to birth.

The history of Silicon Valley is a story of immigration policy colliding with entrepreneurial intent and, more often than not, producing prosperity. The cap tables and patent rolls tell you who built the platforms we use today; the visa rules will tell you who builds the ones we will use tomorrow. If the U.S. wants those platforms to be American, it should stop treating foreign-educated brainpower like a seasonal commodity and start treating it like the strategic asset it is.