PWL Q3 FY26 earnings call.
A source-linked concall view: reported numbers, management language, guidance, commitments and risks that should carry forward.
ConCallIQ research layer
Signal, with the source still visible.
Use the controls below to narrow the view, then follow the evidence into the next layer of context.
Revenue
₹1,082 Cr
verified against source
Revenue YoY
34%
reported change
EBITDA
₹218 Cr
latest reported figure
Source
screener in
record provenance
Actual signal trajectory
Where this quarter sits.
Quarter read
What the record says.
Physicswallah delivered a strong Q3 FY26 with ₹1,082 crore revenue (+34% YoY) and ₹218 crore EBITDA (20.2% margin), growing PAT to ₹102 crore despite ₹23 crore one-time charges from IPO and labor code changes. Online contributes 51% of revenue at 31% growth; offline is 46% at 26% growth. The company raised full-year revenue guidance to 32-35% growth, with 9-month revenue of ₹2,980 crore already surpassing prior-year full-year ₹2,886 crore. K12 expansion is the strategic pivot—₹400 crore allocated to school management, 70 new offline centers planned at ₹200 crore capex. New categories (CLAT, NEPG, IIT JAM) contributed ₹32 crore in 9 months; state boards scaled to 175,000 paid learners (₹22 crore); vernacular reached 94,000 paid students (₹24 crore). AI initiatives (Arbutus, PW Talks) are bearing fruit with in-house proprietary data. Risk: K12 currently contributes less than 1% of revenue with payback cycles of 12+ years, and J/NEET growth has slowed to low-single digits in penetrated markets.
Colored figures show movement against the previous available record.
Guidance to track
- Management guided full-year FY26 revenue growth in the range of 32% to 35%, driven by strong performance in both online and offline segments with continued momentum expected in Q4.
- Full-year marketing spends will be significantly lower than last year's 9.6%, at 8-8.5% of revenue, as Q1/Q2 investments generated higher ROI and marketing leverage will emerge in Q4.
- Company plans to open 70 offline learning centers next year with an outlay of ₹200 crore, continuing the asset-light built-to-suit model for expansion.
- Management anticipates K12 will become a larger business than test prep within 5 years (by FY30), benchmarking against China education companies where K12 is 40% of revenue vs test prep at 30%.
Risks flagged
- JEE/NEET enrollment growth has slowed to low-single digits (3-4%) because the denominator is large and penetration is very high in North India. Analysts questioned whether this pattern will replicate in other categories as they mature.
- K12 currently contributes less than 1% of revenue with payback cycles exceeding 12 years as students stay enrolled from grade 1 to 12. Capital allocation of ₹400 crore (and potentially more) is a long-duration bet with uncertain near-term returns.
- Tech and product team costs increased 40% year-over-year despite overall tech spend as percentage of revenue declining due to scale. AI talent competition could pressure margins as the company invests in AI initiatives.
- Management deflected specific questions about K12 school unit economics, margins, and capital requirements beyond the initial ₹400 crore allocation, citing it as a 'long-term story' without providing detailed financial roadmap.
Key quotes
- K12 will become larger than a test prep business down the line 5 years is what I imagine and if you benchmark any company in China—New Oriental, TAL Edu—all are listed, their K12 business is larger than their test business.
- We generally acquire profitable entities and the range of the multiple we give to these companies is anywhere between 8 to 12 and we believe in earn-out structure where we don't take a short majority. We take anywhere between 25 to 50% equity as the first go and up to a 3-4 year earnout structure for the founders.
- Our AI grader tools are completely designed for Indian test prep kids, completely designed by our own in-house proprietary data, millions of students' daily learning behavior patterns, data points or RAG layer as well. Our small language model operates better than Gemini or GPT-4.
Research modules