A venture firm is now funding a college alternative. On September 22, 2026, Andreessen Horowitz unveiled the Horowitz Andreessen Academy — a $35 million, tuition-free, one-year program for 16-to-22 year-olds, built around hands-on AI and startup work instead of a traditional degree.
Bloomberg reported it as "AI school as a college alternative," aimed at recent high school graduates and college dropouts, meant to "identify and train young talent building artificial intelligence products." a16z general partner Erik Torenberg posted that he'd "wanted to build this for a decade," and that he and academy CEO Gagan Biyani — Udemy co-founder and current CEO of Maven — nearly partnered on it six years earlier.
This is a story explainx.ai is tracking closely: it's the same underlying bet our own live workshop and bootcamp model is built on — that practical, instructor-led AI skill-building beats sitting through generic coursework — just funded at a very different scale.
TL;DR — what the academy actually offers
| Detail | Specifics |
|---|---|
| Funding | $35 million from Andreessen Horowitz |
| Who it's for | Ages 16 to 22 |
| Cost to students | Tuition-free |
| Length | One-year Founding Class starts fall 2027; a two-year version starts 2028 |
| Degree granted | None — explicitly not an accredited credential |
| Led by | Gagan Biyani, Udemy co-founder, CEO of Maven |
| Curriculum | Short courses taught by working tech leaders |
| Work experience | Co-ops at Anthropic, Meta, and Stripe |
| Resources included | $50,000 in compute credits, $5,000 travel budget, housing support |
| How to apply | theacademysf.com — applications open now |
| Early deadline | November 5, 2026 |
Why a venture firm is building a school
The pitch, in Torenberg's own framing, is that college has stopped being the fastest path to building real AI products — and that the people best positioned to teach that skill are practitioners, not tenured faculty. Biyani's background makes the bet legible: he co-founded Udemy, one of the first platforms to prove people would pay for practitioner-taught, non-degree courses at scale, and he now runs Maven, which does the same for cohort-based professional education. The academy is that model pushed further — free, full-time, and aimed at the college-age cohort directly, rather than working professionals.
It also fits a broader pattern explainx.ai has covered repeatedly this year: students choosing AI startups over internships, and a growing argument that traditional software engineering fundamentals still matter even as agentic coding scales. The academy is a structural bet on the same trend — that the fastest-growing cohort of AI builders skews young, self-taught, and increasingly skeptical that four years of tuition is the best route in.
What's genuinely new here versus a bootcamp
Reactions on X ranged from "this is where I'd go at 18" to sharper skepticism about the "world-class professors" framing given the roster is startup operators, not academics. Both reactions are reasonable reads of the same fact: this isn't a bootcamp with better branding, and it isn't a university either. Three things separate it from a typical AI bootcamp:
- Duration and cost structure. Most AI bootcamps run weeks to a few months and are self-funded. The academy is a full year (two-year track from 2028), fully funded by the sponsor rather than the student.
- Direct co-op pipelines into frontier labs. Placements at Anthropic, Meta, and Stripe are named partners, not "career services will help you network" — a meaningfully different level of access than most independent bootcamps can promise.
- Resources beyond instruction. $50K in compute credits and a $5K travel budget go well past what a typical program includes, and point at a curriculum expecting students to actually build and ship, not just complete assignments.
What it doesn't solve, and what any prospective applicant should weigh honestly: no degree means no fallback credential if a student decides the startup path isn't working after a year or two — a real trade-off compared to a college degree's portability across unrelated fields and employers who still filter on degree requirements.
Doing the math on what $35M actually buys
It's worth sizing the commitment against the benefits promised, because "$35 million" sounds large in isolation but says little until it's divided by a cohort. A founding class is typically small — bootcamps and selective fellowship programs (Y Combinator's batches, for comparison, commonly run in the low hundreds) suggest a founding cohort here is more likely dozens to low hundreds of students than thousands, at least initially. Spread across a cohort of that size, $35M easily covers the named per-student resources — $50,000 in compute credits and a $5,000 travel budget alone — plus staff, curriculum development, and housing support, with real room left for the co-op program's coordination overhead. That math is a useful sanity check on whether the offer is sustainable or a one-year marketing splash: a $35M fund spread thin across a large open-enrollment cohort would not stretch nearly as far as the same fund spread across a genuinely selective founding class, which is one more reason to expect admissions to be tight rather than close to open enrollment.
The compute-credit figure specifically is worth pausing on. $50,000 in compute is meaningful money for a student building and training their own models or running heavy agentic workloads — well beyond what a free-tier API budget or a student GitHub Education pack provides — and signals the program expects participants to be doing real technical work, not just prompting existing tools.
How this fits the wider shift in AI education
This isn't an isolated experiment — it lands in the middle of a broader move toward practitioner-led, outcome-focused AI education that explainx.ai's own live workshop model and bootcamp comparisons have been tracking through 2026. The common thread across the academy, cohort-based platforms like Maven, and live-workshop marketplaces is a shared bet that direct access to a practitioner who is actively building teaches AI skills faster than a syllabus written a year in advance by someone once removed from the work. What the academy adds that most of those models don't have on their own is capital — enough to remove tuition entirely and fund the compute and travel that turn a course into an actual building environment.
The risk in that model, worth naming directly: practitioner-led education scales worse than a recorded course, because it depends on the practitioners themselves staying engaged, current, and available. A program funded at this scale can absorb that cost for a founding cohort; whether it holds as the program scales to the two-year track starting 2028 is an open question the announcement doesn't answer.
Who this is actually for
Read the fine print, not just the headline number. This fits a specific profile well and fits others poorly:
- Good fit: a self-directed 16-22 year-old who is already building — shipping side projects, contributing to open source, or running a small startup — and wants funded time, mentorship, and lab access to go deeper, without four years of unrelated coursework standing in the way.
- Weaker fit: someone who hasn't yet decided AI/startups is their path and wants the optionality a broad degree provides, or someone who values the credential itself for reasons outside direct hiring (grad school eligibility, visa sponsorship requirements, family expectations).
- Worth double-checking before applying: what happens to compute credits, travel budget, and co-op access if a student leaves early; whether the "co-op" placements are paid work or unpaid shadowing; and how selective admission actually is once the hype settles — a $35M program aimed at a narrow age band will almost certainly be far more competitive than a typical bootcamp's rolling admissions.
What people are still asking
- "Is this basically a talent pipeline for a16z's portfolio companies?" The program doesn't hide that framing — co-op partners and mentorship are drawn from the AI/startup ecosystem a16z already invests in, so students should expect exposure (and possibly future funding conversations) to skew toward that network rather than being neutral.
- "Why start in fall 2027 instead of immediately?" Standing up curriculum, co-op agreements with named partners like Anthropic and Meta, housing logistics, and admissions for a founding cohort takes real lead time — a year's runway between announcement and first class is typical for a from-scratch program at this scope.
- "Is the November 5, 2026 deadline the only chance to apply?" It's described as an early deadline, implying later admission rounds exist, but the academy's own site (theacademysf.com) is the source of truth for the full admissions calendar — check there before assuming the early date is the only one.
Related on explainx.ai
- Students Building AI Startups Instead of Internships — the trend this academy is explicitly betting on
- How to Become an AI Instructor and Get Paid to Teach Live Workshops — the practitioner-taught model explainx.ai runs today
- Top 10 AI Bootcamps in 2026: Complete Comparison Guide — how funded, structured programs like this compare to self-paid bootcamps
- Andrew Ng: Software Engineering Fundamentals Still Matter in Agentic Coding — the counter-argument on what young builders still need to learn
- Your Job in 2027: How AI Will Transform Every Domain — the career-planning context this decision sits inside
- How to Survive the AI Apocalypse: A Practical Guide — skills that hold up regardless of which education path you pick
Primary source: Erik Torenberg's announcement on X and Bloomberg's initial report, September 22, 2026. Applications at theacademysf.com.
Program details reflect the September 22, 2026 announcement and may change before the Founding Class begins in fall 2027 — confirm current terms directly with the academy before applying. Follow @explainx_ai for updates.