Skip College, Ship Agents: a16z Puts $42M Into the Horowitz Andreessen Academy

The internet made it possible to know anything; AI, the argument goes, makes it possible to do anything. On September 22, 2026, Andreessen Horowitz took that thesis and poured $35 million into it. The venture capital giant announced the Horowitz Andreessen Academy (HAA) — a private, full-time, in-person school in San Francisco aimed at young people coming out of high school, and a deliberate, well-funded shot across the bow of American higher education.

The numbers behind the launch are anything but academic. a16z invested $35 million and incubated the Academy in-house, with founding partners — Anduril, Anthropic, Coinbase, Google, Meta, NVIDIA, OpenAI, Palantir, Replit, and Stripe — contributing another $7 million, bringing total initial funding to roughly $42 million. Techmeme’s coverage notes the program is funded by contributions from OpenAI ($10 million over five years) and Microsoft ($12.5 million), among others. Per reporting from Fortune and the New York Post, every student in the founding cohort receives a package that would make most grad students weep: tuition-free enrollment, roughly $50,000 in compute credits, and a $5,000 travel-and-research budget. When compute is the tuition, the tuition is substantial.

An unaccredited “business,” run like a product

Structurally, HAA is a strange beast. It will technically operate as an independent business rather than a school, and it is unaccredited. Founder and CEO Gagan Biyani — co-founder of Sprig, founder of the Dinner Table media cooperative, and a repeat operator — leads the Academy, which a16z incubated with Marc Andreessen and Ben Horowitz personally backing the effort. CBS News reports the Academy intends to seek regulatory approvals for a full two-year program starting in fall 2028, but it is not waiting for permission to begin: the inaugural one-year Founding Class Fellowship launches in fall 2027 with a cohort of roughly 50 students, ages 16 to 22, living and studying in San Francisco.

The pedagogy rests on three pillars the announcement calls Co-ops, Courses, and Pursuits.

Co-ops are work rotations embedded in the academic year rather than confined to summer internships. Students may rotate across roles and companies at any of the ten founding partners — a list that reads like a league table of the AI economy — or across a broader set of hiring partners. The explicit goal is breadth-first career search: try everything, learn what you like, then go deep.

Courses are taught by what the announcement calls “legendary people” — instructors who have founded unicorn companies, invested in them, or written the textbooks others teach from. Sample offerings include “Origins and Innovations in Digital Computing 1940–2025” and “The Modern Software Developer.” The a16z announcement promises visits from Jensen Huang, Satya Nadella, Travis Kalanick, Mira Murati, Ali Ghodsi, and Patrick Collison, among other luminaries. Alongside the technical canon, the curriculum leans heavily on the “soft skills” courses that business-school students consistently over-subscribe — Interpersonal Dynamics, Paths to Power, Negotiations — treating influence and collaboration as core engineering competencies.

Pursuits are the heart of it: personal projects, startup ideas, quests, and rabbit holes that students workshop with coaches, mentors, and visiting guests. A pursuit might be a hard technical problem that cannot be solved without AI, generating new problem datasets, or even three weeks in another country with the Academy opening doors on arrival. The announcement is blunt about what assignments mean in an agentic era: the best ones are now “problems so hard they cannot be solved without AI.”

The network theory of college, rebuilt in SF

The most interesting part of the HAA pitch is not the compute credits or the celebrity lecturers — it is the theory of what college is actually for. Erik Torenberg’s announcement essay argues that higher education’s most underrated function is being a low-stakes environment to test not just ideas but collaborators. Group projects are how you discover who you work well with; a huge share of the last two decades’ great founding teams came out of exactly those networks. The most talented high school students today, the argument runs, largely lack this — their sharpest peers are scattered across Discord servers, group chats, and hackathon circuits. Real, but thin.

So the Academy’s physical presence in San Francisco is not incidental; it is the product. SF, the essay insists, remains the one place where highly-open, smart, energetic people convene, where cold emails get read and event doors are literally open. The bet is that compressing fifty of the world’s most driven 16-to-22-year-olds into one campus in SoMa — with co-ops at frontier labs and Jensen Huang dropping by — beats four years almost anywhere else.

The catch, and the context

Skeptics will find plenty to chew on. It is unaccredited, so the credential it grants is at best a signal to a hiring network that already helped fund it — the ten founding partners are simultaneously its customers, its curriculum, and its capital. There is an obvious circularity: a school built by VCs whose graduates’ best job option is the VC’s portfolio. And $50,000 in compute credits is generous until you remember that frontier-scale AI research routinely burns that in an afternoon; the credits are a starter kit, not a supercomputer. The one-year fellowship versus two-year-program distinction also remains fuzzy — fall 2027 starts the former, while the latter awaits regulators in 2028.

But context matters more than the caveats. This is the same firm whose founders spent two years arguing that AI renders parts of traditional education obsolete. The Academy is that argument with a balance sheet: not “college is dead,” but “college is beatable for a specific kind of student at a specific moment.” Enrollment is invite-only-adjacent — a selective application for the Founding Class Fellowship — and the target user is the teenager who already ships agents on weekends and finds homework insufficiently hard.

Whether HAA becomes a talent flywheel for the AI economy or an expensive experiment in vibes-based pedagogy, the signal is unmistakable: in 2026, the scarcest resource in tech is not knowledge — it is aligned, ambitious, AI-native collaborators. a16z is now in the business of manufacturing them, fifty at a time, tuition-free.

---

Sources for this article are listed below.