S17.1 · Education, Training & Human Capital
Private K-12 schooling and content, estimated at $663B globally, where AI compresses grading and content cost but not the regulated in-person core.
K-12 education providers are private schooling operators and the instructional content built around them. Research and Markets puts global private K-12 spend at $663B (2026) — a figure that blends tuition with content spend and reads high against a bottom-up count: in the US, roughly 5.5M students attend some 30,000 private schools (NCES), about 9% of K-12 enrollment. The commercial question is where AI lands in the cost stack, and the answer is narrow. Content customization and grading compress; the regulated, in-person instructional core — the part families actually pay for — is largely untouched.
Historical growth is mid-single-digit, with a forward view of ~5-7% resting on two payer developments: US school-choice voucher and education savings account (ESA) expansion, and rising private-school demand in Asia and the Gulf. The voucher dollar matters because it is net-new to operators — public money following the student into a private seat. Revenue concentrates in North America, Western Europe and the Gulf, where expatriate and international schools cluster; Asia carries the most fragmented campus count. Ownership is thin at the top: branded operators — Nord Anglia, GEMS, Inspired, Cognita — hold low single-digit share combined against a long tail of independent and religious-affiliated schools. Curriculum publishers, Pearson and HMH among them, sit upstream; commercial real estate landlords supply the campus footprint. Downstream, families pay directly, with government voucher and ESA programs a fast-growing second payer beside them.
At scale the model earns 20-30% EBITDA margins, but it is capital-intensive — most operators own or lease significant real estate, and returns have to clear that asset base. Durability is enrollment persistence, not a subscription mechanic: a family that enrolls one child tends to enroll the next, and mid-cycle switching is costly to the family as well as the operator. The regulatory gating underneath is revocable, not merely burdensome — school accreditation, curriculum approval, voucher and ESA program eligibility. Any one of them, withdrawn, can eliminate enrollment overnight, which makes regulatory standing as much a determinant of asset quality as the enrollment trend itself.
The adjacencies split by what they do to the revenue line. Supplemental digital curriculum defends the seat — it differentiates the school and lifts retention without adding a payer. Tutoring add-ons for enrichment and downward age extension into childcare add dollars, capturing more of a family's spend across the full span of childhood rather than protecting an existing seat.
Agentic AI's clearest effect is on per-student content customization and grading, both compressing toward zero as AI handles the differentiated assignments, feedback and progress tracking that used to consume teacher time. That is margin, not revenue — the customer and the price are unchanged. What AI does not reach is the regulated in-person core: supervision, safeguarding, and the physical presence accreditation and licensure require. Accreditation status, safeguarding licensure and physical facility access remain the durable moat however good AI tutoring tools become elsewhere in the value chain.
The near term and the long term differ in kind. AI-assisted personalization inside existing schools is already underway and largely additive — better margin, better differentiation, same payer. The consequential shift runs on a longer clock: continued ESA expansion could change who the payer is and how price-sensitive that payer is, a funding-model question on a 2-5 year horizon that matters more to long-run structure than any near-term AI product cycle.