The Global AI Talent War: Who Trains the Engineers and Who Gets Them

Artificial intelligence capability is ultimately people, and those people move. The global distribution of AI talent looks very different from the global distribution of AI education, and the gap between them shapes national outcomes.

Where talent is trained:

Strong technical education is far more widely distributed than AI industry. India produces enormous numbers of engineering graduates. China's universities produce a very large share of the world's AI research output and graduates. Eastern Europe, particularly Poland, Romania, Ukraine and Russia, has deep mathematics traditions. Iran, Turkey, Egypt, Nigeria, Brazil, Argentina and Vietnam all produce capable technical graduates in volume.

Elite research training concentrates more narrowly, in American, Chinese, British, Canadian, European and a handful of Asian institutions.

Where talent ends up:

The flow is persistently toward the United States, driven by salaries that can be several multiples of alternatives, concentration of leading laboratories, and access to compute that researchers cannot get elsewhere.

That last factor is underrated. A researcher's ability to do frontier work depends on access to computing resources available at only a few institutions worldwide, which creates a pull independent of money.

Secondary destinations include the United Kingdom, Canada, Switzerland, Germany, the Netherlands, Singapore, the UAE and Japan, each competing with visa arrangements, research funding and salaries.

The countries that lose:

Canada trains world-class researchers and watches many leave for American salaries, with proximity and easy immigration making the flow frictionless.

The United Kingdom and much of Europe experience the same, with the added difficulty that European salaries for senior AI researchers are substantially below American levels.

Emerging economies lose at every level. Nigeria, Kenya, Pakistan, Bangladesh, Tunisia, Egypt, Argentina and dozens of others train engineers who leave for higher incomes, and the loss is not only individuals but the senior expertise required to build institutions.

Russia lost a very large number of technology workers following 2022, which relocated capability to Armenia, Georgia, Kazakhstan, Turkey, Serbia and the UAE.

What remote work changed:

Remote employment altered the calculation fundamentally. An engineer in Lagos, Buenos Aires, Warsaw or Manila can now earn international salaries without emigrating.

For individuals and their families this is straightforwardly good, raising incomes and keeping people at home. For domestic technology companies it is difficult: they cannot match international salaries and lose staff without those staff leaving the country.

The net effect on national capability is genuinely ambiguous, and it is one of the most consequential open questions for technology development outside wealthy countries.

What actually retains people:

The evidence suggests three things matter: interesting work, access to compute and career progression, in roughly that order alongside compensation.

Countries that have retained talent successfully, Israel through its startup ecosystem, Switzerland through research funding, Singapore through deliberate policy, Poland increasingly through product companies, offer work worth staying for rather than merely competing on salary.

Public compute access is an underused lever. Where governments provide researchers with computing resources they could not otherwise obtain, the pull toward foreign laboratories weakens.

The outlook:

Talent will keep concentrating where compute and capital are, because that is where the interesting problems are.

For most countries, the realistic goal is not preventing emigration but building enough domestic activity that some people stay, some return, and the diaspora remains connected. Several countries have made that work. Most have not tried systematically.