Benched, billed, and counted wrong.
A visa written for a shortage became an industry that treats a worker as inventory — petitioned by the thousand, carried on a bench between billings, and held in place by a green-card queue that can run longer than a career. Most of the desks are dark. A few are lit. The lights are the only thing anyone counts. This dossier counts the rest: what the program actually issues, what it actually pays, and who actually profits.
Every tower was founded by someone who wasn't Indian and is run today by an Indian-born CEO — the founder's flag comes down, India's goes up. It's real and striking, but it's the visible peak of the pipeline, not a plan: the India R&D centers predate these CEOs by 14–20 years.
Every arc is an H-1B approval crossing an ocean. Most run from a single origin. And it was not always so — India was 43% of approvals in FY2004; the concentration was built between 2011 and 2016, and it now stands at 71%.
The H-1B is a child of the boom. Congress created it in the Immigration Act of 1990 with a cap of 65,000, and as the 1990s software surge and the Y2K scramble sent the industry hunting for engineers, the cap was raised twice on a shortage argument — to 115,000 in 1998, then 195,000 in 2000. It was written for a moment. But a visa built for a boom does not end when the boom does: when the dot-com bubble burst, approvals did fall with demand — from 331,000 in FY2001 to 198,000 in FY2002 — yet the cap was cut back to 65,000, approvals soon climbed past it anyway, and the two numbers stopped having anything to do with each other. The boom made the visa. The visa outlived the boom.
The headline cap is 65,000 plus a 20,000 exemption for US master's graduates. It excludes universities, their affiliated nonprofits, and government research bodies entirely — those employers are uncapped. It also excludes every renewal. The number Congress debates is not the number the program issues, and the gap is not small.
The debate is about a number that does not exist. Every argument over "raising the cap" or "holding the line at 85,000" is conducted over a figure that describes a fraction of what the program does. It isn't hidden. It's disclosed, in a report nobody reads.
Body shopping is the industry's own word for it. A staffing firm petitions for workers, places them with client companies, and bills the hour. Between placements the worker is benched — carried, unpaid or underpaid, waiting for the next billing. It is illegal: the required wage is owed whether or not the worker is placed. It is also the model — and it is enforced only rarely. The largest immigration settlement in US history, Infosys\u2019s $34 million in 2013, was for exactly this family of abuse.
Every window is an H-1B approval, filling in the program's real proportions. There are many nationalities in the glass — and it would be dishonest to pretend otherwise about which one fills it. Roughly seven in ten are from a single country.
Registrations for the FY2024 lottery reached 781,000 — and 409,000 of them, more than half, were for people entered multiple times by multiple firms. USCIS moved to one entry per person and added fraud investigations. The next year the multiply-registered pile fell to 47,000; the year after, to under 8,000. It isn't a clean experiment — two things changed at once — but the collapse of the duplicate pile is hard to read as anything but gaming.
Every H-1B petition names a prevailing wage level from I to IV; Level I is the statutory floor — routine tasks, close supervision. In 2009–10, 54% of approved positions were Level I. A program sold as a hunt for exceptional talent was, by its own filings, majority entry-level. That share has since fallen hard — to 16% by 2019 — which is its own admission that the earlier number meant what it looked like.
Green cards are capped at 7% per country of origin regardless of that country's size. For Indian nationals the employment-based queue now runs longer than a working life. The visa is tied to the employer, not the worker — so leaving means starting over. This is what makes the bench possible. A worker who cannot leave is a worker who can be carried.
The people most captured by this program are the people on it. A visa holder in a fifty-year queue cannot change jobs, cannot be laid off without leaving the country, and cannot complain. The bench is not a side effect of the indenture. It is the product of it.
The popular story is that India built its schools to feed the H-1B. The evidence tells a stranger one. In the 1990s the prospect of a US visa did pull Indians into engineering — before India's own software industry existed at scale (software was 2% of India's exports in 1995, 26% by 2000). Then the cap held the door mostly shut. The engineers it turned away stayed home, and that stranded supply became the workforce of the offshoring industry that now takes the work. America capped the visa and built its own competitor. The one route that stayed open is the degree: a US master's roughly doubles your lottery odds, and three years of post-study work authorisation buys about three attempts at it.
In 2020 California's civil rights regulator sued Cisco over caste discrimination between Indian engineers on an American campus — a category US employment law had never contemplated. Seattle banned caste discrimination in 2023. No agency records the caste of a visa holder, so the composition of the program is unmeasurable and this dossier will not guess at it. What is documented is what arrived in the workplace.
India's Employment Visa has its own salary floor (~$25,000) and bars free employer changes — the reciprocity is not symmetric. But the sharper point is duration. Peer regimes tie a skilled visa to one employer briefly: Austria's EU Blue Card converts to portable labour-market access after about 21 months. The H-1B ties the worker until the green card — and for an Indian national, that queue is a lifetime. The length of the tie is the leverage.