Alpha School, an Austin-based private school founded 12 years ago, charges up to $75,000 a year for an education fully piloted by AI. Eight new locations opened in 2025 and nearly two dozen more are planned for the fall in San Francisco, New York, Palo Alto, and Malibu. Tech founder families and venture capital households are lining up.
Key Takeaways
- Alpha School charges up to $75,000 a year for its fully AI-driven school with “learning guides” paid six-figure salaries.
- Eight sites opened in 2025, nearly two dozen more planned for the fall, driven by VC families such as Shaun Johnson’s and public backing from Bill Ackman.
- A Chinese study on 26,000 students found AI-assisted homework was faster and better graded, but exam performance dropped 24%. UC Berkeley reached the same conclusion.
Have an AI Sum Up This Article
ChatGPTA $75K AI School for Silicon Valley’s Kids
Alpha School was founded 12 years ago in Austin, before generative AI became a mainstream topic. Its pitch has evolved with the tech: today, the central platform is an AI tutoring engine that tracks each child’s engagement in real time and adjusts content within the session.
Tuition on the flagship sites reaches $75,000 per year. For comparison, a premium Manhattan private school sits around $60,000, and a Texas public high school sits at zero. Alpha School positions the premium as the price of personalized-at-scale.
Eight sites opened in 2025. Nearly two dozen new locations are on the roadmap for the fall, with openings scheduled in San Francisco, New York, Palo Alto, and Malibu. The geography says everything about the target: US finance and tech hubs, where families have both the budget and the appetite for a school that speaks AI to their kids.
The enrollment list stacks up notable names. VC Shaun Johnson plans to enroll his son. Bill Ackman publicly backs the school. Alpha School spokesperson Anna Davlantes states that “learning guides” (the adults staffing each site) earn six-figure salaries, an elevated standard for the education profession.
The Method: LLMs Front and Center, Human Coaches in Support
Alpha School inverts the traditional school model. The curriculum runs on an adaptive AI that poses questions, times responses, detects disengagement, and adjusts content without human intervention. Students no longer follow a shared program: each one moves at their own pace, on their own subjects, inside a system that recalibrates continuously.
“Learning guides” are not teachers in the classical sense. They don’t prepare lessons, the AI does. Their role: motivational coaching, handling emotional roadblocks, keeping the group cohesion. The role is paid better than a standard teacher role precisely because it requires relational and psychological skills, rarer than mastery of a curriculum.
The format re-shapes the teacher-student relationship. Inside an Alpha School classroom, a child can spend more daily time facing an AI than facing an adult. The AI-to-human time ratio is an explicit choice, and it isn’t hidden. The pitch to parents includes the demonstration: your kid progresses faster because the AI never has to coordinate 25 different paces inside the same classroom.
Technically, Alpha School doesn’t disclose which foundation model powers its stack, but public descriptions point to a mix of frontier models dressed in proprietary pedagogical logic. The stack looks like what prompt-engineering firms sell to enterprises, applied to K-12 students.
Also on Horizon:
- Anthropic Samsung Weigh a Custom AI Chip
- OpenAI Brockman Sees the End of Software Interfaces
- Alibaba Claude Code Blocked for Its Employees
The Mixed Results and the Emerging Education Divide
The pedagogical efficacy debate is open. A Chinese study of 26,000 students showed that homework completed with AI is faster and better graded, but exam performance drops 24% for the same students. UC Berkeley reached converging observations on its own cohorts.
Another signal, 81% of students who use AI over extended periods end up outsourcing their cognitive work rather than learning. The risk is a silent decoupling: strong immediate results, decaying residual capacity. The Alpha School method isn’t immune to this dynamic, it structurally depends on it.
Short-term, Alpha School’s commercial storefront will expand. Each new site in SF, NYC, or Palo Alto is a social proof: the families who build the LLMs are enrolling their kids in schools that learn with LLMs. The narrative closes on itself.
Medium-term, the debate will move to the political floor. A $75,000 AI school isn’t accessible to 99% of US families, and underfunded public schools won’t be able to deploy a comparable stack for years. The AI education divide isn’t a hypothesis: it’s already being built at Alpha School. The wider context helps too: the mixed AI job signals at Ramp and Goldman make the “my family needs AI skills from childhood” pitch far easier to sell.
Follow the story on Horizon.
