A Fairfax-owned small bank that spent FY26 finishing a full tech rebuild while its gold book carried the balance sheet. The March quarter already earns a high-teens return on equity. The bet is that the new core system and spare capital let that stick as the deposit and retail build finally starts. The risk is that the build, promised for years and pushed to FY28 again, stays one year away, and you own a 14%-return bank trading a hair above book.
What it is
CSB is a small private bank from Thrissur, Kerala, in business since 1920 and 40% owned by Fairfax. It runs 862 branches, most of them in the South, and does about ₹85,000 cr of total business. Sub-scale and regional, with one clear specialisation: gold.
The ₹40,359 cr loan book is roughly half gold, a quarter wholesale, and about a tenth each in SME (its "business lending group") and retail. So it is a gold-lending bank with a corporate book and small SME and retail books alongside. Management wants to rebalance to gold 30%, wholesale 30%, SME 18% and retail 20% by 2030, but for now gold runs the show.
Gold is the franchise and the closest thing to a moat. CSB has lent against gold in the South for a century, so it originates cheaply through a dense branch network in the gold heartland, underwrites at a low ~67% loan-to-value, and takes almost no credit loss. A large slice is agri gold, which sits outside the 75% LTV cap and earns well. Portfolio yield is around 12%. A real edge, but a narrow one: gold lending is crowded with Muthoot, Manappuram and every other bank, and CSB's gold growth has come more from the rising gold price than from new customers, with account count actually falling.
Deposits are the weak spot, and the whole case sits here. CASA is ~20% and drifting down, about half of term deposits are expensive bulk money, and NRI deposits are ~13%. There is no liability moat: CSB has no deposit pricing power and pays up to fund itself. For years the old core system could not launch competitive deposit products, so the franchise never got built.
That is why the bank earns a mid-teens ROE (14% in FY26) rather than the high-teens its gold economics could support. Two things cap it: the costly deposit base, and a fat capital cushion (CRAR ~20%) that is not yet fully worked. Returns are decent, not elite.
Ownership and management are strengths. Fairfax is a patient, aligned owner. Pralay Mondal, from Axis, Yes, HDFC and Standard Chartered, has run the bank since 2022, and the board is full of career bankers. They have delivered on their near-term calls. The one thing they keep promising and deferring is the deposit and retail build.
Why now
The tech rebuild is done. New core plus ~50 surround systems went live in a year from May 2025 with almost no disruption. Their own reason for a decade of weak retail was that the old system could not support products or a sales team. That is gone, and product launches have started.
The run-rate is ahead of the reported year. FY26 ROA was 1.29% and ROE 14.14%, held down by a soft Q1. Q4 printed 1.53% ROA and 17.66% ROE. If that is the new base and not a one-off, the bank already earns more than it is guiding to.
On where ROE lands: guidance is ~1.5% ROA and ~15% ROE into FY27. That 15% is conservative and capital-heavy. It can go higher for two reasons. Cost-to-income is 62% and is meant to fall from FY28 as the tech and branch spend earns out, which lifts ROA. And the bank holds far too much capital (CRAR 20.7%), so as it grows and uses that up, ROE rises on its own. Together, high teens (17 to 20%) is realistic by FY29-30 if the build works. Q4's 17.66% shows it is not a fantasy. It all turns on deposits and retail converting.
FY26 actual | Q4 FY26 exit | FY27 guided | FY30 if build works | |
|---|---|---|---|---|
ROA | 1.29% | 1.53% | ~1.5% | ~1.5 to 1.65% |
ROE | 14.14% | 17.66% | ~15% | high teens |
NIM | 3.76% | 3.83% | 3.75 to 4.0% | 3.75 to 4.0% |
Cost-to-income | 62.5% | 61.9% | 60 to 65% | low 50s |
Loan growth | 27% | ~25% | mid 20s, easing |
Triggers
Already working
Q4 asset quality best of the year, SMA book lowest in five quarters, credit cost down to 0.18% for the quarter.
Moving gold up-ticket lifted portfolio yield ~40 bps as customers refinanced in from costlier NBFCs.
Deposit and fund costs fell through the year, sharpest in the bulk bucket.
Possible from here
Cost-to-income breaking below 60% from FY28. The single biggest lever.
Deposit repair: new CASA products, a TASC team helped by a friendlier LCR rule, a Dubai office for NRI money once West Asia settles.
Retail assets turning positive from FY28, off the deposit base rather than a book-first push.
Spare capital used up as growth compounds, lifting ROE with no change in ROA.
Free options
A large AFS bond book taken when RBI changed the rules, which went against them as yields rose. If yields fall it swings back into treasury gains.
A working-capital-against-gold product for SMEs, targeted at 5% of the book by 2030, higher-yielding and stickier than plain gold.
Risks
The build slips again. The main risk. The 20% CASA, the bulk-deposit reliance, and the retail timeline pushed to FY28 are the same problem in different clothes, and this exact promise has a long record of delay. If it slips, ROA stays 1.3 to 1.4%, spare capital sits idle, and the high-teens ROE never shows.
Half the book is one asset. Credit is safe, LTV covers a 10% price fall. But two years of book and earnings growth leaned on the gold price. If gold flattens, the growth stops before diversification is ready.
NIM has more downside than up. As the mix shifts off high-yield gold and CASA stays low, margin sits at 3.75 to 4.0%. That puts the whole returns story on cost, which puts it on execution.
Credit cost normalises up. FY26 was flattered by running down unsecured and pausing SME. As both resume, provisions rise off a low base. Some Q4 strength was migration-related technical defaults reversing, which does not repeat.
AI cyber overhang. An RBI advisory dated 27 April wants an AI cyber-risk plan at the board by end-June. Management sees no clarity yet on cost, and would not rule out that it delays the cost-to-income drop.
What the price expects
1.25x book, under 10x earnings, near the 52-week low of ~₹275 against a high of ₹574.
Book value is ₹282 and compounds at close to the ROE, because the bank keeps almost all its profit and pays next to no dividend. At 14 to 15% ROE that is mid-teens book growth every year. Hold the multiple flat and you earn roughly that. The argument is all in the multiple.
At 1.25x the market pays a thin premium to book. That is the price of a bank stuck at a 14% return with a weak, bulk-heavy deposit base and growth flattered by the gold price. Baked in: more of the same, mid-teens book compounding, no re-rating, no credit for the 17.66% Q4 or the 2030 plan.
If the build lands, ROE settles high teens, book compounds nearer 17%, and the multiple has room toward 1.8 to 2x, where small private banks earning high-teens returns on a real deposit franchise trade. You get the faster book growth and the re-rating together. If the build stalls and ROE fades to 12 to 13% with capital idle, book compounds slower and the thin premium goes, dragging the stock back toward book.
Three outcomes. Hold 1.25x and compound book mid-teens: ~15% a year on doing what it already does. High-teens ROE and a re-rate toward 2x: low-to-mid 20s. Stall to book value: mid-single-digit with a de-rating drag. The price sits on the first and gives you the second for free. It turns on one question: does the deposit and retail engine start, or does next year stay next year.
What would change my view
Negative read. CASA slips below 20% and bulk stays near half of term deposits into FY27. Retail is still flat or shrinking a year out and FY28 gets nudged again. Cost-to-income does not break below 62% by end-FY27. Gold flattens and growth stalls before the rest of the book is ready. Any of these makes the high-teens ROE a permanent 13 to 14%, and at that return 1.25x book is about right, not cheap.
Positive read. CASA and NRI deposits turn up, bulk reliance falls, funding cost drops. Cost-to-income falls in FY27 not FY28, pulling ROA above 1.5% early. Non-gold lending grows visibly, proving the bank is more than a gold book. Spare capital gets used and ROE prints high teens for a few quarters running, not one. On this bank, going from a believed 14% return to a believed 17% is the difference between a thin premium to book and roughly double it, on a book compounding mid-teens underneath.
TenetFour Research. Educational analysis only. Not investment advice or a recommendation to buy, sell, or hold any security. Author may hold a position. Readers are responsible for their own decisions.