The $1 Billion Bet That Teachers Are Optional

A Stanford dropout who built the first billion-dollar AI company is now applying the same playbook to classrooms, betting a fortune that a sub-$1,000 tablet can replace teachers for a billion children.

By Carry and Conquer Publications

Joe Liemandt spent 25 years avoiding journalists. The man who took Trilogy Software from a Stanford dorm room to the cover of Forbes, then quietly assembled ESW Capital into an acquirer of more than 100 enterprise software companies, resurfaced in 2025 with a different pitch entirely. He has pulled $1 billion out of his own companies to fund Alpha School, a K-12 network where roughly a dozen U.S. campuses have already stripped teachers out of the classroom, and he is now telling anyone who will listen, including foreign ministers of education, that within five years a tablet costing less than $1,000 will teach every child on the planet everything they need to know in two hours a day.

The SaaS Playbook, Applied to Second Grade

Liemandt's core thesis is not really about pedagogy. It is about labor substitution, the same move that built his fortune. Trilogy and ESW made money by using software to automate work that used to require people; Alpha School applies that lens to the one profession Silicon Valley had mostly left alone. Students spend two hours a day working through adaptive software such as IXL, Lexia, and NoRedInk, moving forward only after clearing a 90% mastery threshold. Adults are not called teachers. They are called guides, and their job is explicitly not instructional. Guides earn above-market pay, reportedly over $100,000 a year in some markets, on a 5-to-1 student ratio, and their function is limited to motivation, dashboard-checking, and life-skills coaching in the afternoon block that follows the academic session.

Alpha reports its results through NWEA MAP Growth, a third-party standardized test, and claims students post roughly 2.3 times the annual academic progress of national norms, with some cohorts reaching 99th-percentile scores and mid-year data showing conditional growth percentiles in the 83rd-to-99th range across grade levels. These figures come from Alpha's own published analyses rather than an independent audit, a distinction the network's critics keep returning to. Two investigative reports, one from WIRED in October 2025 and a second from 404 Media in February 2026, found that a substantial share of the "guides" fielding student help requests in 2023-24 were remote workers based in the Philippines and Colombia, employed by Liemandt's other software ventures rather than credentialed as teachers. The same reporting described default webcam monitoring that persists when students take devices home, and internal flags describing some AI-generated lesson plans as doing "more harm than good."

Pricing a School Like a SaaS Tier

Tuition at Alpha's campuses runs on a ladder that mirrors enterprise software pricing more than traditional private-school economics. The flagship Austin campus and most locations sit in the $40,000-to-$65,000 range; Palo Alto and other high-cost markets reach $75,000. At the opposite end, the Brownsville, Texas campus charges $10,000, a price point Liemandt's team has repeatedly cited in white papers and in charter-school applications as evidence the model works across income levels, even as WIRED's reporting documented multiple Brownsville families withdrawing their children. Layered beneath the private-school tier is TimeBack, a $15,000-tuition microschool product aimed squarely at families eligible for state vouchers, structured so that a roughly $12,000 voucher brings a parent's out-of-pocket cost down to the $300-to-$400-a-month range Liemandt describes as the eventual mass-market price.

That pricing ladder only works if regulators let the underlying model touch public money, and there the results have been mixed. Unbound Academic Institute, an organization affiliated with Alpha's founders, filed cyber-charter applications in Pennsylvania, Arizona, North Carolina, Arkansas, and Utah. Arizona approved its application; Pennsylvania rejected it in January 2025, with the state's Department of Education calling the instructional model "untested" and finding it failed to show alignment with state academic standards. Arkansas and Utah also denied the bid. Pennsylvania's staffing plan for the proposed school, which would have served 500 students, listed one head of school, five senior guides, ten junior guides, and one special-education guide, all operating without a physical campus, with state testing to occur in a rented space and a mailing address at a commercial shipping location. Wikipedia's entry on Alpha School also documents a web of interconnected for-profit vendors, including 2 Hour Learning, Trilogy Enterprises, and Crossover Markets, contracted by the nominally nonprofit schools, with board members at Unbound Academy affiliated with those same vendor companies.

Why the Window Is Open Now

None of this has slowed Liemandt's timeline. Alpha is expanding into roughly a dozen new campuses across Arizona, California, Florida, New York, North Carolina, and Virginia, with cities including Chicago, Atlanta, and San Francisco added to the map for the coming school year. The expansion is landing at the exact moment voucher programs in Texas, Florida, and Arizona are scaling up statewide eligibility, which is the regulatory tailwind Liemandt needs to move TimeBack from a boutique private-school product into something closer to a subsidized mass-market offering. Every other well-capitalized entrant into K-12, whether a curriculum publisher, a charter operator, or a traditional edtech company, is still pricing and structuring itself around a teacher-centered classroom. Liemandt is the only major backer explicitly building for a post-teacher unit economics model, which means the "$300-a-month AI school" category currently has almost no competition defining its standards, its regulatory treatment, or its ceiling on scale.

The bet is enormous and the proof is still thin. Alpha's growth numbers come from a test the company chooses to administer and interpret, cyber-charter regulators in four of five states have already said no, and the two most detailed outside investigations of the model found remote non-teachers, surveillance-by-default, and lesson plans flagged internally as unreliable. Liemandt is nonetheless the first person with both the capital and the enterprise-software instincts to treat those as solvable bugs rather than reasons to slow down, and he is moving to lock in the pricing and distribution model before anyone else, regulator or competitor, gets to define what an AI-taught education is allowed to cost.