India's Edtech After Byju's: What 2026 Looks Like
Byju's collapse cleared the table. The companies that survived — and the new ones building on top of the wreckage — are doing different work.
In 2021, Byju’s was valued at $22 billion. By the end of 2023, the company was in financial distress. By 2024, insolvency proceedings had begun. By 2025, the brand had been carved up — pieces sold, pieces shuttered, the founder and the leadership team gone. The largest edtech failure in any major market in the last decade was an Indian one.
It is also a useful clearing event. The narrative that consumed Indian edtech from 2018 to 2022 — aggressive sales-led acquisition, premium tutor brands, multi-thousand-dollar bootcamps for school children, hyper-funded subscriber acquisition — is gone. What replaced it, by 2026, is a much more interesting picture.
What killed Byju’s, briefly#
The collapse was the result of three compounding mistakes.
Aggressive sales tactics produced regulatory backlash. Parents were aggressively up-sold into multi-year contracts for primary-school children, often with loans they did not understand. Consumer protection regulators began action; the Karnataka government’s order in 2023 marked a turning point.
Capital allocation across acquisitions destroyed value. Aakash Educational Services, Toppr, Whitehat Jr., Epic, and several others were acquired at peak prices and integrated poorly. Whitehat Jr. in particular — coding-for-kids — had aggressive marketing under fire and was effectively shut down post-acquisition.
Revenue recognition issues created a credibility cliff. The Deloitte audit issue in 2023, the delayed financial filings, and the eventual lender actions cascaded into the collapse.
These are corporate-finance failures, not edtech failures. The product question — does online education work? — was never really answered by Byju’s. It was answered by the companies that survived.
What survived and what is working#
A non-exhaustive map of the active companies in 2026:
Physics Wallah (PW) has emerged as the largest profitable edtech company in India. Founded by Alakh Pandey, originally a YouTube educator, PW priced its courses at 5-10x lower than Byju’s and Unacademy and built a profitable business on volume. They now serve test prep (NEET, JEE), CA, GATE, government exams, and the K-12 boards. Their model is closer to traditional coaching at internet scale than to the premium subscription model Byju’s pursued. As of 2026 they have 2,500+ offline learning centers (Pathshalas) across India, complementing the online product.
Unacademy is the second large surviving brand, though leaner than at peak. Their pivot to test-prep and away from K-12 has been the right strategic move. Their AI tutor product, Cosmo, launched in 2024, has been a meaningful product win.
Vedantu, after layoffs and restructuring, has refocused on small-cohort live-tutoring and AI-powered homework help. Less ambitious than the peak strategy, more sustainable.
Cuemath, focused specifically on math instruction for K-12, has had a quieter but more durable trajectory. AI-augmented one-on-one tutoring has emerged as their core product, and the engagement metrics are strong.
Skill-Lync and Newton School for engineering and computer science upskilling. Targeted at college graduates and early-career professionals.
upGrad continues as the largest in the executive education segment — MBAs, data science programs, executive certificates, partnerships with international universities.
Sleepy Owl for early childhood is a smaller but interesting play.
Khan Academy India, DIKSHA (the government’s national K-12 platform), NPTEL (the national IIT-led MOOC platform) all serve substantial free-tier audiences and shape the broader ecosystem.
The AI tutoring shift#
The most consequential 2024-2026 shift is AI-driven tutoring. Several specific applications have crossed the “useful enough to charge for” threshold.
Homework help and doubt solving is the most penetrated use case. A student photographs a problem, the AI explains the solution step by step. Quality has improved enormously with frontier models; Cuemath, Vedantu, and a long tail of smaller players have shipped credible products. Average revenue per user is low but the unit economics work because there is no human tutor cost.
Personalized practice generation is the second use case. Given a student’s recent performance, the AI generates practice problems calibrated to their level and tracks improvement. PW’s product team has built this aggressively.
Conceptual explanation in regional languages has been a quiet success. A Tamil-speaking class 10 student can ask a chemistry question in Tamil and get a fluent response — something not possible at this quality two years ago. This is the Indic LLM opportunity in education.
Voice-based learning for non-English-medium students is emerging. Sarvam’s voice models, in particular, have been integrated by several edtech platforms.
Tutor augmentation rather than replacement is the dominant pattern. A live tutor with an AI co-pilot — generating examples on the fly, surfacing student misconceptions, drafting parent updates — produces better outcomes and better economics than either pure-AI or pure-human tutoring.
The interesting unsettled question is whether AI tutoring at scale can match the outcomes of an excellent human tutor with rich student context. The honest answer in 2026 is “approaching, not yet.” But the gap is closing.
The K-12 segment shape#
K-12 is the largest segment by audience but the most operationally complex.
The boards and curricula — CBSE, ICSE, the state boards — all changed substantially with the NEP 2020 (National Education Policy) rollout. The shift toward competency-based education, foundational literacy and numeracy emphasis, and a less rote-heavy approach has changed what edtech content looks like.
Test prep continues to be the most economically valuable segment — NEET, JEE Main and Advanced, CUET — because parents will spend on outcomes. The competitive intensity here is high; PW has set a low price floor and the competitors have had to follow.
Supplementary content — adaptive practice, video explanations, doubt-solving — is the largest engagement category. Increasingly commoditized; the differentiation is quality and regional language coverage.
School solutions (selling to schools rather than to parents) is the segment most edtech companies under-invested in. Schools themselves are increasingly integrating digital learning tools, often through state-level initiatives.
The structural challenges#
Three honest challenges remain.
Outcomes measurement is hard. Without rigorous, third-party-validated learning outcomes data, the marketing claims of edtech products are difficult to evaluate. The Karnataka government’s 2023 actions explicitly cited this; the broader regulatory direction is toward outcome substantiation.
Affordability and the digital divide. A child in a rural Bihar village without a smartphone or stable internet does not get access to the best edtech, even when it would help most. The government’s DIKSHA platform and the various state-level initiatives are designed to bridge this; the bridge is not yet complete.
Teacher capacity is the binding constraint at scale. AI tutoring can supplement but does not yet fully replace good teaching. India’s structural teacher shortage — particularly in math, science, and English — limits how far the model can scale.
What outsiders should learn#
For edtech investors and operators outside India, three observations carry across:
-
Hyper-funded, sales-led, premium-priced edtech does not work at scale in price-sensitive markets. The Byju’s playbook failed not because the product was bad but because the price was wrong for the market it was sold into. Other markets — Indonesia, the Philippines, parts of LATAM — should learn from this.
-
The AI inflection is real. Tutor augmentation is reshaping the unit economics and the product surface. Companies that have not figured out their AI-tutor strategy in 2026 are behind.
-
Outcome accountability is coming. The next regulatory wave in edtech globally will be about substantiating claims. Companies that built around outcome data — Cuemath’s math-improvement metrics, PW’s exam-rank tracking — are positioned to comply; the rest will scramble.
Where pdpspectra fits#
We build platform and AI engineering for education companies out of our Kathmandu and Boston offices. Our work spans content delivery infrastructure, AI tutoring system integration, learning analytics platforms, and the engineering work that makes regional-language content delivery operationally sustainable.
Related reading: the AI in education tutoring and admin post, the Indic LLMs post, and the school ERP migration checklist for Nepal.
Indian edtech is past its first cycle. Talk to our team about what comes next.