FanYu Enters Global Unicorn Database as AI Education Shifts to 'Action-Level Planning'
Beijing Duwo Technology, operator of the children's learning brand FanYu, has been named a sample enterprise in the UP Global Unicorn Enterprise Database, according to the UP Global Unicorn Enterprise Data Analysis Report 2026 released on September 22 at a conference in Qingdao. The report was jointly issued by the Unicorn Engineering Institute, Renmin University of China's private enterprise research center and Tsinghua University's new quality productivity institute.
Entry criteria include a valuation above RMB 7 billion (about US$980 million), proprietary or disruptive technology, a hard-to-replicate business model, roughly ten years since founding, policy alignment and no major negative publicity, with a valuation reference date of August 1, 2026. Founded in 2015 by Huang He, a former founding product partner at Toutiao, FanYu reports more than 100 million registered users and over five million paying users across 328 Chinese cities, plus operations in 25 countries and regions in Southeast Asia, North America and the Middle East. The wider report counted 1,881 unicorns globally worth RMB 60.14 trillion (about US$8.4 trillion) across 14 sectors, 52 countries and 360 cities.
Huang said the basis of competition in education is shifting from course content, teaching talent and traffic to learning data, user profiles, long-term feedback and dynamic learning-path management.
FanYu's in-house "Harness" AI system pairs a cross-disciplinary knowledge graph with an AI diagnostic brain covering in-school exams, classroom learning, after-class practice and online answering. When a student gets a question wrong, it traces prerequisite knowledge points, weighs submission timing against a historical mastery curve, and identifies whether the cause is conceptual confusion, weak reading or carelessness before adjusting subsequent teaching actions. A "nano-level" knowledge tag framework decomposes points to common-sense cognition and stratifies learners into struggling, average and advanced paths.
The industry reading is that value is moving from content distribution to service delivery: a continuously running diagnosis, planning and feedback system rather than videos and question banks. That implies higher fixed research and teaching investment but lower marginal service cost, and moves the competitive bar from who holds more content to who can adjust teaching actions dynamically.
Reading Camp is the fastest-growing line, with eight levels and 24 graded reading sets tracking the national curriculum revision that reframes English as language competency built on heavy input and output rather than exam drilling. After new textbooks took effect in 2024, live-stream channels scaled quickly; as reform awareness spread to lower-tier cities in 2025, sales multiplied several times over 2024. The company says Reading Camp tops single-session course sales on Douyin, WeChat Channels and Xiaohongshu, with live-streaming above 80% of the product's sales.
On cost, FanYu injects teaching materials, past exam papers and teacher explanations into the system, with Tsinghua and Peking University teachers recording lessons and provincial subject leaders tuning the AI, expecting fixed costs to dilute as the paying base and renewals grow. Huang framed competitiveness as being engaging, effective and efficient — a child finishing in 15 minutes what took 30 — while arguing for restraint in AI, grounding answers in official materials to avoid the hallucination common to general-purpose models.
Policy is the largest variable ahead. In April 2026 the Ministry of Education and four other departments issued the "AI Plus Education" action plan calling for personalised learning and equitable access, and in September Beijing released updated AI-in-education guidance emphasising personalised learning.
FanYu says it will keep iterating its system and explore an "AI plus human teacher" model, aiming to make personalised learning an everyday product rather than a luxury. Whether AI education reaches scale depends on closing a loop of low ticket price, high retention, verifiable outcomes and profitable scale.