Microsoft, Alphabet and several of the labs driving the artificial intelligence boom are run or shaped by people born and educated in India. India itself ranks fourth globally on AI performance, behind only the United States, China and Singapore, according to the State of India’s Digital Economy 2026 report from the ICRIER-Prosus Centre for Internet and Digital Economy. It accounts for 26 percent of global AI users. It attracts about 1 percent of global private AI investment. And it has recorded one of the steepest net outflows of AI talent of any major country.

By several measures, India has more AI talent than anywhere else. It also keeps less of it at home. The reasons go beyond the usual shorthand of “brain drain” to a set of structural factors: money, computing power, customers and credentials. Together they have pulled talent out of the country and, until recently, given it little reason to return.

A Talent Leader That Keeps Losing Talent

India leads the world in AI skill penetration. AI-related skills show up in Indian professionals’ LinkedIn profiles at three times the global average, the highest rate of any country, according to the 2026 Stanford AI Index. The same Stanford report draws on data from the talent-analytics firm Zeki, which ranks India second in the world for total AI authors and inventors in 2025, with 50,460 researchers. The country trails only the United States, at 220,520, and edges out Germany, at 48,520. The ICRIER-Prosus SIDE 2026 report independently reaches the same conclusion, naming India home to the world’s second-largest AI talent pool after the U.S. Indian students adopted generative AI tools faster than American students did, with usage climbing sharply over two years.

The Stanford report, now in its ninth year, recorded a net outflow of AI talent from India, with a 2025 net migration figure of about minus 16.9 per 10,000 professionals, the steepest of any major country. So the country with the most AI skill per capita has also been losing AI talent on net.

The picture is shifting, though. The report itself describes India as “transitioning from being a net exporter to a net absorber” of top AI talent, a shift it ties to conditions in the United States. The number of AI researchers and developers moving to the U.S. has fallen 89 percent since 2017, including an 80 percent drop in the past year alone. India is still losing talent on net, but by the report’s account the reason it is keeping more is less that home has improved and more that the U.S. has become harder to reach. That tension, a large historical outflow that is now narrowing, sits underneath the rest of this story.

Where the Money Is

Capital is the most powerful factor, and the gap is wide.

AI startups in the United States raised about $121 billion across 765 funding rounds in 2025. Indian AI startups raised about $643 million across 100 deals in the same period, according to figures from the startup tracker Tracxn that were shared with TechCrunch. That is a difference of roughly 188 times. Over a longer span, India drew about $11.1 billion in cumulative private AI investment between 2013 and 2024, rising to roughly $12.3 billion when government funding is included, according to the Stanford AI Index 2025 as reported by Business Standard. That is far behind both the U.S., at $471 billion, and China, at $119 billion, over the same period.

The funding gap shows up most in pay. Compensation data from Levels.fyi, last updated May 31, 2026, shows OpenAI research scientist packages ranging from $771,000 for an L4 to $1.47 million for an L5, with the highest reported package at $1.9 million. The Wall Street Journal reported that the company’s average stock-based compensation across its roughly 4,000-person workforce reached about $1.5 million per employee in 2025. The average AI researcher in India earns about 1.7 million rupees, or roughly $20,000, according to self-reported data on Payscale. Even strong early-career researchers earn a fraction of what a comparable hire makes in the San Francisco Bay Area.

The gap also shapes how companies behave. Large Indian firms, including some with market values that rival global peers, have historically spent far less on research and development than their international counterparts. A researcher who wants to work on hard, open-ended problems rather than services and implementation has had few well-funded options at home.

Data-Rich, Compute-Poor

Modern AI depends on access to specialized hardware, mainly high-end graphics processing units, or GPUs. India has long been on the wrong side of that divide.

Analysts call India’s position a “data-rich, compute-poor” paradox. The country generates a large share of the world’s data but has historically owned a small fraction of the world’s AI computing capacity. A widely reported February 2025 Motilal Oswal report concluded that India “faces challenges such as a lack of homegrown large-scale AI models, dependence on foreign technology, and limited AI hardware infrastructure,” citing a shortage of high-end AI hardware, limited access to advanced GPUs and cloud computing, and insufficient research funding. The foundational technology that Indian startups build on is, by most measures, imported rather than homegrown.

That matters for where talent goes. Training a frontier model requires tens of thousands of GPUs running for months. Where that infrastructure does not exist, the most ambitious work in the field cannot be done locally, so researchers go where the machines are.

This is the area changing fastest. India’s government launched the IndiaAI Mission with a budget of about 10,300 crore rupees, or $1.25 billion. The GPU rollout has passed its initial targets. From a goal of 10,000 GPUs, the country had deployed more than 38,000 by early 2026, with another 20,000 announced at the India AI Impact Summit in February 2026, bringing the total to more than 58,000. The GPUs are made available to startups and researchers at 65 rupees per hour, well below global market rates. In December 2025, three U.S. tech giants committed major sums to Indian AI and cloud infrastructure within days of one another. Microsoft pledged $17.5 billion over 2026 to 2029, its largest investment in Asia. Amazon committed more than $35 billion through 2030. And Google announced a $15 billion AI data hub in Andhra Pradesh, its largest such project outside the United States. Reliance, for its part, has announced plans for a Jamnagar data center that Bloomberg reported could reach 3 gigawatts at a cost of $20 billion to $30 billion, which would rank it among the world’s largest. The infrastructure is no longer static. But the talent that left did so during the years when it was.

Why the Founders Leave Too

Money and machines do not fully explain why founders, not just job-seekers, leave.

In January 2026, the Economic Times reported, with corroboration from Business Standard, that more than 100 Indian AI startup founders had relocated to the United States or planned to, most to the San Francisco Bay Area. Companies that have already moved their headquarters include Composio, Smallest.ai, Beatoven.ai, Meetstream.ai and GetCrux. Founders cited reasons beyond funding: faster access to enterprise customers who buy AI products, larger pools of specialized talent to hire from, and an established network of accelerators, investors and other founders.

There is also a feedback loop around credibility. Indian startups often gain standing at home only after they succeed abroad. A founder who raises money in Silicon Valley is taken more seriously in India than one who never left. That pushes patent filings, headquarters decisions and long-term value out of the country, and it encourages the next group of founders to follow.

Commentators point to a structural risk: if the pattern holds, with India using foreign AI at the application layer while the underlying technology is built elsewhere, the country could end up supplying data and usage fees to foreign systems while the high-value work happens overseas. Some have called this the danger of becoming a “digital colony.”

A Pipeline Built to Export

Part of this predates AI. India’s elite engineering pipeline, anchored by the Indian Institutes of Technology, has functioned as a talent export system for decades, and the data shows that the strongest students leave at the highest rates.

A peer-reviewed study by Prithwiraj Choudhury, Ina Ganguli and Patrick Gaulé, “Top Talent, Elite Colleges, and Migration: Evidence from the Indian Institutes of Technology,” published in the Journal of Development Economics in 2023, tracked the highest scorers on the IIT entrance exam. Among the top 1,000 scorers, 36 percent had migrated abroad. Among the top 100, the figure rose to 62 percent. Among the top 10, it was nine out of 10. The higher the score, the more likely the student was to leave.

The system selects for problem-solving under tight resources, the same trait Silicon Valley rewards, and then watches its best students leave for graduate school in the U.S., where India has overtaken China as the largest source of international students.

The result shows at the top of the industry. Alphabet’s Sundar Pichai came through this pipeline at IIT Kharagpur; Microsoft’s Satya Nadella took a different route, through Manipal Institute of Technology and then U.S. graduate schools. Among AI-specific leaders, Aravind Srinivas, the Chennai-born, IIT Madras-educated founder of Perplexity, and Arvind Jain, the IIT Delhi graduate who founded the enterprise AI company Glean, are both products of the IIT system. The talent is real, and it is running much of the industry from outside India.

The Doors Abroad Are Closing

For most of this period, the United States was an easy destination. That is changing, and it is the most volatile part of the picture.

In September 2025, the U.S. administration imposed a $100,000 fee on H-1B visa petitions, a sharp increase from the prior range of roughly $1,700 to $4,500. After early reporting described it as an annual charge, the White House and U.S. Citizenship and Immigration Services clarified within days that it was a one-time fee applying only to new petitions filed after Sept. 21, 2025, not to renewals, extensions or current visa holders. The policy falls hardest on India. Indian nationals account for about 71 percent of all H-1B recipients and more than 80 percent of computer-related H-1B positions, according to Fortune’s reading of USCIS data. For many employers the visa, long the main route for skilled Indians into U.S. technology jobs, became far more expensive, about 10 times the prior cost per applicant, immigration attorneys told CBS News.

Other pressures point the same way: reduced federal research funding, the rollback of earlier AI safety programs and a more restrictive stance on immigration. Together they have made the U.S. less attractive than it was two years ago. A Richmond Fed economic brief predicted that one likely result of the fee will be U.S. firms offshoring operations to countries such as Canada or India, citing 2023 research showing that earlier H-1B restrictions had already pushed companies to relocate R&D-intensive jobs abroad. The effect, in other words, could be to move the very work that talent used to migrate for.

Analysts now discuss the prospect of a “reverse brain drain,” with skilled Indians choosing to stay or return and strengthening hubs such as Bengaluru, Hyderabad and Pune. Some state governments, notably Tamil Nadu, have launched programs to draw researchers back. Major AI firms, including OpenAI, Anthropic, Google and Perplexity, have moved to set up or expand an Indian presence. Anthropic opened a Bengaluru office and named the former Microsoft India managing director, Irina Ghose, as its local head. India is now the second-largest market for some of these companies’ products.

Whether this is a turning point or a pause is an open question. The pull factors that drew talent abroad are narrowing but have not closed, even as the reasons to stay home and the obstacles to leaving both rise.

For most of the past decade, every incentive pushed the same way. The money was larger abroad, the computing power existed only abroad, and the customers, the graduate schools and the founder networks were all there too. The visa door was open. That pattern says less about Indian talent than about where the global AI economy put its resources. Talent went where the resources were, and India produced so much of it that the industry would look very different without those who left.

The next phase is uncertain. India is spending heavily on computing power, global capital is flowing into its data centers, and the traditional destination is making entry harder. The country spent a decade as the world’s largest exporter of AI talent. The question now is whether it can keep that talent, or bring it home.