While the rest of tech tiptoes around immigration politics, Nvidia is hiring foreign talent like it's 2021 — and paying them like the AI boom depends on it.
The Summary
- Nvidia increased H-1B visa certifications 20% year-over-year (1,200 in first half of fiscal 2026 vs. 1,000 last year) while Google dropped 57% and Amazon fell 30%
- Software engineers command up to $391,000 base salary, research scientists $356,500, before equity or bonuses
- CEO Jensen Huang, Taiwan-born, calls immigrants "crucial to the company's mission" as competitors quietly retreat from visa sponsorships
The Signal
The H-1B data tells a story about competitive moats that transcends politics. Nvidia isn't just bucking an immigration trend. It's demonstrating that when you're genuinely supply-constrained on specialized talent, you pay market price and deal with the paperwork. Software engineers earning $391,000 base isn't charity. It's what happens when AI infrastructure becomes the most valuable substrate in computing and the number of people who can design CUDA kernels at scale fits in a small conference room.
The timing matters. While Amazon and Google pull back H-1B hiring by 30-57%, Nvidia is expanding 20%. This isn't about immigration philosophy. It's about strategic necessity. Training an LLM is different from optimizing ad delivery. The former requires deep expertise in distributed systems, GPU architecture, and numerical precision that maybe 10,000 people globally understand at a production level. The latter? You can hire from a much deeper pool.
"Foreign hiring has slowed across tech, with the Trump administration's immigration crackdown prompting some firms to reduce H-1B visa sponsorships."
But here's what the salary data actually reveals:
- Hardware roles pay slightly less than software, suggesting the constraint is algorithmic innovation, not chip design
- Research scientists top out at $356,500, lower than engineering, meaning Nvidia values deployment over pure R&D
- Manager and director roles pull comparable numbers, implying flat hierarchies where technical work commands premium compensation
The equity component, unstated in these filings, completely changes the equation. Nvidia stock is up roughly 850% over three years. A $250,000 stock grant in 2023 is worth over $2 million today. That's not compensation. That's generational wealth accumulation in a single vesting cycle.
What Nvidia is actually doing is arbitraging global talent markets. Pay $391,000 to someone in Bangalore or Shanghai who could start a well-funded AI company anywhere, give them equity that makes them a millionaire in four years, and extract the institutional knowledge required to maintain a 90% GPU market share in AI training. The H-1B visa isn't immigration policy. It's a retention mechanism.
The Implication
Watch who's still hiring internationally when everyone else is "rightsizing for efficiency." That's your signal for which companies believe their technical edge is non-replicable and which are managing cost centers. If you're an AI engineer deciding where to build your next three years, follow the H-1B money. The companies still paying premium rates and navigating visa complexity are the ones building infrastructure that compounds, not features that ship.
The broader play: Nvidia isn't just hiring people. It's accumulating institutional knowledge about how to make AI systems actually work at scale. That knowledge walks out the door the moment you stop paying market rate. Google and Amazon can afford to slow hiring because they have depth. Nvidia is still building the airplane mid-flight.