a16z launches a tuition-free school for 50 young people: less a university alternative, more a talent pipeline for AI

By CBS News · September 22, 2026.

Andreessen Horowitz, one of Silicon Valley's most influential venture capital firms, has unveiled the Horowitz Andreessen Academy, an unaccredited school conceived as an alternative to university for young people who want to found AI companies. The one-year, tuition-free program will begin operating in the fall of 2027 with just 50 places, as the firm itself has confirmed.

The curriculum does away with exams and homework: students will learn by designing AI systems, raising capital, selling companies and storytelling (the ability to tell and sell an idea), all through real projects rather than closed courses. The faculty includes Sam Altman, OpenAI's chief executive, as an instructor or guest speaker. The academy is aimed at recent high school graduates or those who have completed up to two years at an "elite university": it is not a program open to anyone, but a filter applied to a profile that the traditional education system has already preselected.

The most revealing part is not the teaching model but who funds it: $42 million contributed by Anthropic, Google, Meta, OpenAI, Nvidia and Palantir, according to a16z's own figures. These are companies fighting tooth and nail for leadership in AI models, yet here they coincide in putting money into the same talent pipeline. Andreessen Horowitz, which in August had closed a $1 billion fund for physical AI infrastructure, also plans to extend the program to two years from 2028, with tuition "similar to that of elite private universities" — that is, it will stop being free as soon as it proves it works.

Our reading: despite the headline, this is not a mass alternative to university. With 50 places a year and entry requirements that already filter for the best students, the Horowitz Andreessen Academy is, above all, an early selection mechanism: identifying future founders before the university system absorbs them, training them in the culture and tools of a very specific ecosystem, and having them already embedded in it —as future customers, partners or portfolio startups for the investors themselves— by the time they need to raise funding. That Anthropic, OpenAI and Google, head-to-head rivals in the model arena, coincide here as funders points to something we have commented on before: the real competition is no longer fought only on benchmarks, it is also fought over who controls the talent pipeline and the training standards of the next startup ecosystem.

The experiment also puts figures on an increasingly serious debate: the value of the traditional university in the age of AI. Ben Horowitz puts it bluntly: the education that worked for the Industrial Revolution does not work for the AI revolution. There is truth in that —the pace at which tools change renders much of the content of a university degree obsolete before it is finished— but it should be read with caution: no one is better positioned to sell the idea that university is no longer needed than someone who needs young, cheap and immediately available engineers and founders.

In the short term the risk is real and should be stated plainly: an unaccredited qualification is a high-risk bet for any young person without a pre-existing safety net, and it only works if the surrounding ecosystem —funds, contacts, visibility— provides an outlet even without a piece of paper to show. For the 50 chosen it will probably work: they will have access, mentors and capital that the vast majority of university students will never see. For the general debate over whether a university degree is needed to work in AI, this program provides no generalizable answer: it provides an extreme and very well-funded case.

In the long run, however, the underlying instinct —that education has to reinvent itself for an economy in which AI absorbs the routine and what remains is judgment, creativity and the ability to build— connects with the thesis we hold about the future of work: the winner is whoever learns to orchestrate AI, not whoever memorizes how things were done before it existed. If the abundance AI promises truly arrives, the relevant question will not be whether a university degree was needed, but whether millions of young people —not just 50 a year in San Francisco— had a real opportunity to learn to build with these tools.