PhysicsWallah pulled a sharp U-turn on its K-12 strategy, announcing it will go "100% asset-light" and avoid M&A. This reverses its Q3 FY26 plan to make K-12 bigger than test prep, now pivoting hard into AI-driven learning with in-house small language models. The company expects these new AI tools to generate "meaningful revenue" starting in FY27.
How We Got Here
PhysicsWallah's CFO had pitched K-12 as a future growth engine in Q3 FY26, expecting it to surpass their core test prep business within five years. This sudden shift means no capital deployment for school acquisitions, despite those earlier ambitions.
The Numbers
- PhysicsWallah is building small language models (SLMs) in-house, claiming 5X speed and 1/10th cost compared to frontier models.
- Co-founder Prateek Maheshwari states their AI models, trained on "billions of data points" from 3.5 million students, can outperform some frontier models in accuracy.
- The upcoming Aryabhata 2.0 (math solver) and AI Tutor (problem-solving, memory) are slated for FY27 rollout.
- PW's ARPU for online courses is under Rs 4,000, driving the focus on cost-effective AI tutors at Rs 10 per day.
- Currently, PW's AI suite includes Ask AI (3.05 million doubts solved) and AI Guru (99.45 million queries resolved).
What Happens Next
🇮🇳 Why This Matters for India
For students in Tier-2 and Tier-3 cities, PW's affordable AI tutors could significantly improve access to personalized education at Rs 10 per day.
The Take
This is a clear bet on margins: scaling AI is cheaper than scaling physical schools, especially for a company used to sub-₹4,000 ARPU. PW understands that affordability, not just cutting-edge AI, is the real frontier for edtech in India.
Source:
MediaNama ↗