Paid AI Fellowships in 2026 and What the Jobs Actually Pay
Updated June 14, 2026
The biggest names in AI have started paying people to learn the job. As of June 2026, Anthropic, OpenAI, and Google all run paid fellowships and residencies that hand promising researchers and engineers a real stipend, a compute budget, and a direct line to the people building frontier models. The money is serious. So is the competition.
Anthropic's Fellows Program pays about $3,850 a week, roughly $15,000 a month, plus another $15,000 a month in compute, with workspace in Berkeley or London. OpenAI runs a near-identical Safety Fellowship and, separately, a Residency that pays residents like full-time staff, around $18,300 a month for six months. Google's AI Residency is a twelve-month research training role. Same shape as Meta's trades academy, just pointed at AI research instead of HVAC: a big company paying to mint the talent it needs.
So the fair question is the same one we asked about the trades. Does the work these programs lead to actually pay enough to build a life on, and where does that money go furthest? Here are the numbers, straight from federal wage surveys, plus an honest read on who gets in and what the catch is.
What these AI careers pay
Median annual wages from the U.S. Bureau of Labor Statistics, 2025 data. Median means half earn more and half earn less, so read it as the middle of the road, not a ceiling. The range shows where entry-level pay starts and what the top tier earns. None of these include the equity that frontier labs throw on top.
The stipend is big. So is Bay Area rent.
A $15,000-a-month fellowship sounds life-changing until you price a one-bedroom in Berkeley or San Francisco. That is the quiet catch with every big-lab program: the money is concentrated in the most expensive housing markets in the country. The good news is the career it feeds is not. A computer research scientist or ML engineer earns six figures in Austin, Raleigh, or Columbus too, where the same paycheck buys a lot more life. Before you chase a coast, see where the salary actually goes furthest.
The programs worth knowing about
These are the paid, credible paths into frontier AI work. Most are competitive and most reopen on a rolling basis, so the move is to know them now and watch for the next window.
- Anthropic Fellows Program. A four-month empirical research fellowship, about $3,850 a week plus a heavy compute budget, based in Berkeley or London. Over 40 percent of the first cohort were hired full-time afterward. The clearest example of a lab paying to grow its own researchers.
- OpenAI Residency. Six months as a full-time, fully paid employee, around $18,300 a month, embedded with research teams and with a real shot at converting to a permanent role. The highest cash of the bunch.
- OpenAI Safety Fellowship. A stipend-funded research program, $3,850 a week plus compute, running September 2026 into February 2027. Easier to enter than the residency, more research-first.
- Google AI Residency. A twelve-month research training role aimed at jump-starting a machine-learning research career. Longer runway than the others.
- Independent fellowships (MATS, Constellation's Astra, and similar). Outside the big labs, several nonprofits run paid AI-safety research fellowships that act as broader on-ramps, often friendlier to people switching in from another field.
Programs are listed for convenience. AffordMap does not run, endorse, or have any affiliation with these fellowships, and the wage figures above are not a promise of pay. Application windows and stipends change, so confirm details on each program's official page.
Get started: skills and jobs
Whether you are building toward a fellowship or skipping it and going straight for the job, these get you from reading to applying.
Some links on this page are affiliate links. If you click through and take action, AffordMap may receive compensation at no cost to you. This never influences the salary data, cost-of-living figures, or calculations shown, all data comes from U.S. government sources. See our full disclosure.
Common questions about paid AI fellowships
Are AI fellowships actually paid, or is it a foot-in-the-door thing?
They pay, and the numbers are not small. The Anthropic Fellows Program and the OpenAI Safety Fellowship both run a $3,850 weekly stipend, roughly $15,000 a month, plus about $15,000 a month in compute and direct mentorship from staff researchers. OpenAI’s Residency goes further: residents are full-time employees earning about $18,300 a month for six months. These are not unpaid internships dressed up with a nice title.
Do you need a PhD to get into one?
No, and that is the part people miss. The big labs recruit strong engineers, physicists, mathematicians, and self-taught researchers, not just AI PhDs. Anthropic says outright that it wants people who can do good empirical research even without a prior background in AI safety. A residency or fellowship is often the bridge for someone with real technical chops but the wrong line on their resume. The bar is demonstrated ability, not a specific degree.
What does an AI researcher make after a fellowship?
The fellowship is the on-ramp, not the destination. The job most fellows are aiming at, Computer and Information Research Scientists in federal data, pays a median of $140,300 a year nationally. Entry-level sits near $82,200, and the top 10 percent clear $230,630, before the equity that frontier labs hand out. Figures are from the U.S. Bureau of Labor Statistics, 2025 data.
What is the difference between a residency and a fellowship?
A residency usually means you are a full-time employee for a fixed stretch, paid like staff and embedded on a real team. OpenAI’s Residency is the clearest example: full salary, six months, with a path to a permanent role. A fellowship is typically a stipend-funded research program, four to six months, where you are not a regular employee but you get money, compute, and mentorship to work on a hard problem. Residencies pay more cash up front. Fellowships are easier to enter and lean more research-first.
What is the catch?
Three of them, honestly. They are brutally competitive, they are short, and they sit in the most expensive housing markets in the country, Berkeley, San Francisco, and London. A $15,000-a-month stipend reads very differently once Bay Area rent takes its cut. And there is no guarantee of a full-time offer at the end, though the odds are real: over 40 percent of Anthropic’s first fellowship cohort were hired full-time afterward. Go in for the work and the network, not a promised job.
When do the 2026 cohorts open?
The marquee 2026 windows have mostly closed: Anthropic took applications for its May and July cohorts through January, and OpenAI’s 2026 Residency round is shut. OpenAI’s Safety Fellowship runs September 2026 into February 2027. The practical move is to track the next openings rather than scramble when one drops, the labs reopen these on a rolling basis and rarely give much notice.
Can you build an AI career without getting into a big-lab program?
Yes, and most people do. The occupations these programs feed, software development, data science, and machine-learning engineering, hire across the whole country, not just at five frontier labs. A fellowship is a fast track, not the only track. Strong ML and Python certificates plus a portfolio of real projects will get you into the same job market, and you can do that work in a metro where your salary actually stretches.
Get told when the next AI cohort opens
The labs reopen these fellowships on a rolling basis and rarely give much warning. Drop your email and we will send a note when a new paid AI program shows up, no spam, unsubscribe anytime.
Eyeing a hands-on path instead of a research one? See the free trades-training programs and what those jobs pay.
Source data is published by the U.S. Bureau of Labor Statistics and other federal agencies. AffordMap is not affiliated with or endorsed by any government agency. Per BLS policy, "BLS.gov cannot vouch for the data or analyses derived from these data after the data have been retrieved from BLS.gov."