Andreessen Horowitz puts $35 million into a college alternative for 50 students a year -- economists say the model that works for 50 won't work for 5,000

Andreessen Horowitz is putting $35 million into a bet that a hand-picked group of teenagers doesn't need college at all. The venture firm is the lead investor behind the Horowitz Andreessen Academy, a new unaccredited, for-profit school opening in San Francisco that will take about 50 students for a free, one-year Founding Class Fellowship starting fall 2027 -- no grades, no tests, no degree, and admission based on “proof of work: what they have built, shipped, or earned” rather than transcripts.

The full raise is $42 million, combining a16z's $35 million with undisclosed individual amounts from seven named backers: Adam D'Angelo, Tobi Lütke, Tony Xu, Fidji Simo, Garry Tan, Joe Liemandt and Shyam Sankar. Udemy co-founder Gagan Biyani is CEO; Marc Andreessen and Erik Torenberg join him on the board. Students keep 100% of any intellectual property they create, and a16z takes no equity -- the firm's return, if there is one, is reputational and relational rather than a stake in what students build.

One outlet's account differs here in a way worth flagging directly: CBS News described the $42 million as coming from the ten founding companies themselves -- Anthropic, Google, Meta, OpenAI, Nvidia and others. TechCrunch, Fortune and the San Francisco Standard instead consistently name a16z plus the same seven individual investors as the funders, describing the ten companies as non-financial founding partners supplying mentors, speakers and computing infrastructure rather than cash. Three independently reported, internally consistent accounts naming specific individuals outweigh one outlet's shorthand, so that's the version this piece uses.

“The training that worked for the Industrial Revolution isn't going to map perfectly onto the AI revolution.” — Ben Horowitz, co-founder, Andreessen Horowitz

The free year is the pitch; the paid one is the actual business model. By fall 2028, the academy plans to launch a two-year program with tuition “similar to an elite private university” -- reporting puts that in the $60,000-to-$90,000-a-year range, roughly Stanford's own sticker price. That program remains unaccredited, meaning no federal student loans, no Pell Grants, and no path to aid for a family that can't pay list price. Biyani has framed the project as building on top of higher education, not replacing it: “we believe in college,” he told Fortune, positioning the Academy as innovation rather than a rejection of the existing system.

Two economists who study alternative credentials aren't convinced the second year works. Bryan Caplan of George Mason University's objection is about scale: something that can work for 50 hand-picked people isn't going to work even for the top 5,000, because the extreme selectivity that lets a non-degree credential overcome employer skepticism is exactly what makes it non-replicable past a tiny cohort. Ryan Craig, managing director at Achieve Partners, targets the financing instead: “it's a bootcamp with a plan to be a college, but I'd say not a realistic plan,” he told Fortune, adding that “it's hard to compete with free money from the government” -- a reference to the string of prior for-profit alternative-college attempts that charged private-university tuition without federal aid access, and failed.

The ten founding partners -- Anthropic, OpenAI, Google, Meta, Nvidia, Anduril, Coinbase, Palantir, Replit and Stripe -- are supplying speakers, mentors and, per multiple accounts, compute infrastructure, not tuition dollars. That's a cheap commitment for companies of that size to make, and a real one for a 17-year-old choosing between a guaranteed Stanford acceptance and a free year at an untested school with no track record of graduates.

The academy isn't the first attempt to build a credential around “proof of work” instead of a transcript -- Y Combinator has run on a version of that logic for two decades, and Peter Thiel's Fellowship paid teenagers $100,000 each to skip college outright starting in 2011. What's different here is the scale of the ambition: Thiel's program never tried to become a two-year institution charging tuition, and Y Combinator never positioned itself as a college replacement at all. The Academy is explicitly trying to be both a Thiel-style talent bet in year one and an accredited-feeling institution by year two, on the theory that AI-company mentorship and a residential cohort can substitute for what a university actually sells: a legible, transferable credential an employer outside the ten founding partners' own network will recognize on sight.

That's the part neither Caplan nor Craig's critique fully resolves and the Academy hasn't yet had to answer: whether “built something Anthropic's mentors were impressed by” travels as far outside Silicon Valley's own AI-company network as a degree does. The Founding Class won't finish anything until 2028 at the earliest -- which means the institution's actual test, the one both economists are pointing at, doesn't arrive until the tuition bill does, and doesn't fully resolve until the first cohort tries to get hired somewhere that isn't one of its own ten founding partners.