Technology

Andreessen Horowitz Plans a Tuition-Free Startup School in San Francisco

Martin HollowayPublished 2w ago3 min readBased on 4 sources
Reading level
Andreessen Horowitz Plans a Tuition-Free Startup School in San Francisco
source:a16z.com

Andreessen Horowitz is creating the Horowitz Andreessen Academy, a private, full-time, in-person school in San Francisco for young people who want to build or join a Silicon Valley startup. The firm says it is a school, not a fund or accelerator, and the first one-year program is scheduled for fall 2027. a16z

The funding totals $42 million led by Andreessen Horowitz. The Verge The project also raised an additional $7 million from its founding partners. New York Post

Those founding partners double as the launch employment network. The Academy will launch with 10 partners: Anduril, Anthropic, Coinbase, Google, Meta, NVIDIA, OpenAI, Palantir, Replit, and Stripe. The cohort is small by design. The program is full-time and in-person.

Admissions targets recent high school graduates. The first program will accept only 50 students and will be free. Applications are opening for a tuition-free fellowship. San Francisco Chronicle The Academy plans to offer a two-year program at a later date.

There is no degree or accreditation. There are no traditional grades, tests, or homework. Instruction centers on short classes led by tech figures including OpenAI CEO Sam Altman, paired with co-ops that place students in roles at tech companies. The model is project work, time on live systems, and direct access to working operators, without the credential layer.

Students must move to San Francisco and arrange their own housing for the one-year program. Students are promised more than $50,000 in compute credits, which is prepaid time on cloud computers for building with AI, plus a $5,000 travel and research budget from Andreessen Horowitz.

In my view, the decision to bundle compute, co-ops, and co-location instead of a standard curriculum is the point. The credits remove the cost limit for training models, running inference at scale, which means serving an AI model to many users at once, or shipping prototypes that need GPUs, the specialized chips used for AI work. The co-ops give access to production systems, structured code review, and on-call discipline for keeping software running. Having raised two children as school shifted toward work, I have seen credentials matter less when tools change fast. If these 50 place well, founders gain junior builders who already know deployment workflows. If not, the cost is a 50-person experiment. What this enables is a parallel track where practical experience sorts talent more than examinations.