Innovation: Artificial Intelligence Talent (Private Sector)

Sector: 9% of GDP (Technology) 
Estimated Savings: $15B over 10 yrs / .24% of federal budget 
Political Challenge: High  

I. Overview 

America’s edge on artificial intelligence talent — that is, the people behind the technology — is eroding, endangering our economic, national security, and innovation leadership. Protecting our leadership means modernizing our visa system to better retain top STEM talent. Beyond visa reform, a wider set of reforms and investments are needed to equip American workers and the heartland for success (at best, unprecedented success) in a new world.  

II. Problem 

America’s shortfalls in AI top talent run deep.^1,2 A wealth of literature suggests that the global talent behind artificial intelligence is some of the brightest and most entrepreneurial in the world, and more of it would fortify America’s economic, security, and innovation leadership. Salary bidding wars for machine learning PhDs offer some proof of the idea that our AI ecosystem is bottlenecked by scarce talent.^3,4 While rising worker compensation is something to celebrate, it could be rising even faster — driven by higher innovation rather than lower competition — without those bottlenecks.⁵ That would be to the public’s benefit. 

These talent shortfalls won’t necessarily resolve themselves. Typically, high salaries would gradually draw new workers into the sector. But universities, unable to compete with frontier labs’ attractive compensation offers, struggle to retain or hire faculty to expand AI programs. It’s a known problem in economics, sometimes termed “eating the seed corn.” Industry pulls educators into corporate labs, leaving fewer qualified to train the next generation.⁶ Moreover, policy barriers to high-skilled immigration prevent us from recruiting that talent from overseas.  

With a stronger base of top talent, the private sector could:  

• enable further “vertical” breakthroughs in what AI can do 

• support “horizontal” deployment into the broader economy, for example improving healthcare and education 

• spark growth in more places in our country 

Current policies neglect foreign-born, home-grown talent. The average AI researcher working at an American company was trained here but born abroad.⁷ This means that, in effect, our graduate education system and our visa system are our talent policy. But our schools are significantly more open than our workforce, and much of America’s full potential — in growth, innovation, and even national security — is lost in this gap.

Think of it this way: our graduate education system is like a fertile valley where crops from every corner of the world have taken root and flourish side by side. Then ask: When summer comes, what should we harvest? That’s the question answered by our visa system.  

We’re getting it wrong. Annually capped visas like H-1B and Green Cards sort through high demand in fairly arbitrary ways, like lottery-style distribution and per-country limits. Uncapped alternatives like O-1 and J-1 visas go underutilized due to ambiguous criteria. The resulting “general climate of uncertainty, complexity, and restriction” creates a revolving door, with significant and rising shares of international graduate students trained in U.S. universities forced to return to competitor countries like China.^8,9 While the Biden Administration took meaningful steps to streamline these processes and expand visa eligibility,¹⁰ it did not enact the full extent of possible reforms, and executive and congressional action is still sorely needed.  

AI talent policy is an opportunity for regional development. AI firms are highly concentrated around a handful of coastal metros, particularly California’s Silicon Valley and Virginia’s Loudoun County. The burgeoning AI sector has the potential to worsen regional inequality — or, with the right policy interventions, spark new talent clusters and inclusive growth across the country. 

For the most inclusive growth, invest in the wider workforce. According to our best forecasts, “unprecedented economic growth will only be captured by the countries whose workers are prepared in data and AI basics,” and current skills gaps promise to supercharge inequality.¹¹ As ever with rapid technological change, individual and national success hinge on the education of our workforce. And much like Sputnik spurred the National Defense Education Act, international competition including from China’s DeepSeek must provoke new action to build the AI workforce.  

III. Recommendations 

Visa reforms are critical. The Administration can take some actions unilaterally, for example:  

• extending O-1 visa renewals and expanding their eligibility,  

• promoting the use of O-1s and J-1s by industry through soft power,¹² and 

• developing accelerated Green Card pathways through the Department of Labor.¹³  

Congress must make broader reforms to support STEM talent, and the lowest-hanging fruit include removing Green Card country caps and prioritizing visas based on earning potential or earning history rather than expanding programs wholesale. A new "national security visa" could be designed to prevent illicit technology transfer and inhibit China’s AI ecosystem while cultivating national talent.¹⁴  

For national greatness and inclusive growth, invest in a broader base of talent. Congress should leverage community colleges first and foremost for expanding AI education.¹⁵ Congress could fund a new program to issue moderate annual grants to a wide swath of community colleges for new AI offerings. Executive agencies like the Departments of Education and Labor, and the National Science Foundation, can engage in civil society partnerships and competitive grantmaking to develop curricula and skills-based certifications, especially for the crucial middle tier of AI managers and implementers.¹⁶  

To solve the teacher shortage, a wide range of training and upskilling programs are plausible, including a national emerging technology educator challenge, updates to the criteria of key federal education grants to include artificial intelligence applications, and a new Digital Frontier Teaching Corps.^17,18  

Any of the above investments could be geographically targeted to particularly support regional tech hubs and growing AI clusters, from Ann Arbor to Atlanta.¹⁹ Moreover, expansions of visas can be tailored to support heartland regions, for example through a Heartland Visa.²⁰  

IV. Risks and Politics 

Congressional or executive action on immigration reform faces headwinds. The administration and much of the GOP caucus has a strong restrictionist mandate. Success likely requires either presidential endorsement and attention, support from influential immigration advocacy groups, and / or championship by key senators.²¹  

But several factors could enable targeted reforms. The Republican caucus has shown some openness to net-zero immigration proposals that expand some programs while reducing others. High-skilled immigration is much more popular than immigration broadly — including with President Trump. A potential Supreme Court decision on DACA could effectively force congressional bargaining.  

Any reform must also thoughtfully address legitimate national security concerns about technology transfer risks, which could be heightened by expanded visa programs but reduced by better visa design.²² An expanded national security visa program could present a particularly flexible and effective policy response to these risks.