← all companies · 2026-07-07 · moat: narrow · raw .md

Axis Bank: The Year Profit Almost Vanished

An anomaly in column two

Open this bank's ten-year accounts and read the profit column aloud, in crores of rupees: 7,450 … 8,358 … 3,967 … 464 … 5,047 … 1,879 … 7,252 … 14,207 … 10,919 … 26,492 … 28,191 … 26,548.

Stop at the fourth entry. In the year ended March 2018, Axis Bank — then already one of India's largest private banks, with ₹46,614 crore of revenue, thousands of branches, and tens of millions of customers — earned a net profit of ₹464 crore. Not 464 crore of loss; something stranger. A machine of that size ran flat out for a full year, collected every EMI and every fee, paid every salary, and kept, at the end, roughly one rupee of every hundred it took in. Earnings per share: ₹1.78. Two years earlier the same machine had earned ₹8,358 crore. Two years later it would nearly vanish again — ₹1,879 crore in March 2020.

An anomaly like that is a gift to the curious reader, the way a bent light ray was a gift to physicists: it means some large, invisible force is operating, and the anomaly is your instrument for measuring it. The force in this ledger is leverage — the defining physics of banking — and no Indian bank's published numbers demonstrate it more instructively. This chapter uses Axis Bank's own accounts as the laboratory: first to derive how leverage works, then to ask the only question that matters about a bank that has already shown you its failure mode — whether the repair is structural or merely recent.

The amplifier: leverage derived from first principles

A bank's balance sheet has a peculiar architecture. Against every rupee of the owners' capital, it holds roughly nine or ten rupees borrowed from others — mostly depositors. This is not recklessness; it is the design. The bank's product is intermediation, and intermediation only pays if you run a large book of other people's money over a thin sliver of your own.

But study what that architecture does to arithmetic. The bank earns its living on the whole pile of assets — a good bank might clear about one rupee of profit per hundred rupees of assets after all costs. One percent. A pathetic margin, until you remember whose money the hundred is: since only ten of those rupees belong to the owners, a 1% return on assets becomes roughly a 10% return on equity. Leverage multiplies the sliver into a living.

Now run the machine in reverse, because the amplifier has no rectifier in it — it cannot tell gains from losses, and multiplies both identically. Let loans go bad equal to just 2% of assets. Against total assets that is a stumble; against the owners' ten-rupee sliver it is a fifth of all the capital, gone. The first line of defence is a year's operating profit — the earnings the machine throws off before providing for losses — and this is exactly what the ₹464 crore year shows: revenue grew from ₹41,409 crore in March 2016 to ₹46,614 crore in March 2018, a healthy 13% expansion of the machine itself, while profit fell 94%. The engine ran fine. Its entire output, essentially every rupee of pre-provision profit, was shovelled into the furnace of loan-loss provisions against corporate credit extended years earlier. When a year's output isn't enough, the furnace starts on the capital itself — and then the bank must go to shareholders, cap in hand, for more.

Which is precisely the next thing the ledger records. Watch the equity capital line (face value ₹2 per share, so this line counts shares): ₹479 crore in March 2017, ₹513 crore in 2018, ₹564 crore in 2020, ₹613 crore in 2021 — from about 240 crore shares to about 306 crore, nearly 28% more shares in four years, sold to plug the holes the loan book had burned. Dilution is how a bank pays for old mistakes with its future: every new share entitles someone else to a slice of all subsequent recoveries. Eleven years on, the arithmetic still bears the scar. Net profit rose 3.6 times between March 2015 and March 2026 (₹7,450 crore to ₹26,548 crore), but earnings per share rose only 2.7 times (₹31.42 to ₹84.89) — the missing quarter is the permanent toll collected by crisis-era share issuance. The dividend tells the same story in a sadder voice: skipped or token for most of a decade (zero in 2018, 2020 and 2021; around 1% of profit in recent years, a 0.07% yield), because a bank rebuilding capital cannot afford generosity.

That is the whole science, demonstrated on real money: leverage converts small errors of judgment into existential accounting, and the bill is paid in provisions first, capital second, and shareholders' ownership third.

One more turn of the lens, because the same arithmetic explains where banking profit actually settles. If leverage multiplies whatever judgment it is given, then the durable winners in this industry are decided not by who lends — everyone lends, and in booms everyone lends to the same people at the same inadequate prices — but by two quieter variables: who funds most cheaply, and who errs least. Cheap funding widens the spread being multiplied; scarce error keeps the multiplication working in the right direction. Every strategic decision Axis has made since 2019 is legible as an attempt to move those two variables, because its own ledger had just demonstrated, at a cost of several years' profit, that nothing else matters.

A bank named twice

How did this particular institution come to run the experiment? Axis was born in 1993 as UTI Bank, one of the first private licences of the liberalisation era, sponsored by India's giant state-run mutual fund — an origin that left it with a lasting oddity: no promoter in any meaningful sense. Even today the "promoter" holding is just 8.14%, a residue of state-linked institutions rather than a founding family; foreign and domestic institutions own most of the bank (and swapped roles lately — foreign holdings fell from 52% to 42% between June 2023 and March 2026 while domestic institutions rose from 30% to 43%). In 2007 the bank shed its parental name and became Axis.

Ambition was the house style. Through the 2000s boom, Axis — like its development-finance-descended peers, and unlike its most conservative ones — lent enthusiastically to infrastructure and industrial projects. The 2010s delivered the verdict via the regulator's asset-quality reviews, and columns two through six of our ledger record the sentence being served: 2017, 2018, 2019, 2020 — four years in which a franchise of national scale earned, cumulatively, less than it had earned in the single year 2016. New leadership arrived in 2019 with a mandate written by the ledger itself: recognise everything, provision conservatively, rebuild capital, shift the book toward retail granularity, and get the deposit engine — always this bank's relative weakness against the giants — pulling harder. In 2023 the bank bought Citibank's India consumer business, acquiring at a stroke the affluent-customer franchise it had struggled to grow (the ₹10,919 crore profit of March 2023, the ledger's last dip, is largely that acquisition's one-time accounting charge, a purchase expensed rather than a wound).

Since then, the column reads like a different institution: ₹26,492 crore, ₹28,191 crore, ₹26,548 crore. Return on equity, which averaged just 11% across the ten years — the crisis years drag like an anchor — has run around 13–16% in the recent three.

What it costs to be third

Replicating Axis Bank would cost what replicating any large Indian bank costs: a licence earned over years, roughly five thousand branches' worth of physical trust, decades of deposit relationships, and the risk-and-compliance machinery of a systemically supervised institution. That barrier is real and keeps the club small; India has produced only a handful of private banks at this scale in three decades. Revenue of ₹1,32,538 crore in March 2026 — nearly four times the ₹35,727 crore of March 2015 — is the size of the machine a challenger would need to match before the contest even began.

But Axis also illustrates the hierarchy inside the club. The third- largest private bank earns structurally less than the top two, for one deep reason this chapter has already met: funding cost. The cheapest deposits flow to the banks whose names are synonymous with impregnability, and a bank that spent 2017–2021 in the newspapers for bad loans and capital raises pays a trust-tax on every rupee it gathers. The moat around the industry is high; the moat around third place is lower, and made of the very discipline whose absence the ledger records.

The Munger movement: apologising with other people's shares

Munger's first instrument, as always: incentives. A bank with no promoter is run by professional managers on professional horizons, and the ledger you have just read is what that can produce when the horizon is short — growth booked today, provisions discovered by one's successors. The repair since 2019 has visibly lengthened the horizon. But Munger would insist on the distinction between a reformed institution and a reformed era: the true test of a credit culture arrives only in the boom, when bad lending looks like alpha and the board is asked why growth lags the ambitious. Axis has now compiled three consecutive years of ₹26,000-crore-plus profits. It has not yet compiled the harder exhibit: a full euphoric cycle declined.

Invert the question — what must not happen this decade? — and management's assignment writes itself:

Never again pay for mistakes with new shares. The 28% dilution of 2017–2021 is the single most expensive number in this chapter. The decade's first commandment is to carry enough capital, priced into ordinary profitability, that the next downturn is absorbed by earnings rather than by shareholders' ownership. Equity capital has crept from ₹613 to ₹622 crore since 2021 — near-zero issuance. Keep it that way through a full cycle and the record becomes an argument.
Close the deposit gap. The bank's profitability ceiling is set by its funding cost. The Citi franchise brought exactly the right customers; the task is to make granular, low-cost deposits grow faster than the loan book for ten straight years — dull, decisive work.
Make the payout a signal. A ~1% payout ratio a decade after the crisis tells owners the balance sheet still isn't trusted by its own stewards. A steadily rising dividend would be the cheapest credibility the bank can buy.
Institutionalise the credit veto. The 2019 repair is associated with specific leadership. Convert it into structure — outcome-linked incentives for credit officers, a risk function that reports past the growth function — before the people who remember 2018 retire.
Let the return on equity, not the loan book, be the scoreboard. Between 11% (the ten-year average, scars included) and the recent 13–16% lies the entire difference between mediocrity and a compounding machine at 9× leverage.

Owning the amplifier

Now the Buffett exercise: buy the whole bank for about ₹4.17 lakh crore and pocket last year's ₹26,548 crore — a starting earnings yield above 6%, the market charging just 15.8 times earnings and about 1.9 times the ₹692 book value. That is a visibly cheaper ticket than India's premier franchises command, and the discount is not a market error; it is a precisely worded opinion. It says: this machine's average output over long periods, including its accidents, has been ordinary — an 11% ten-year return on equity — and you may have the upside of reform in exchange for bearing the risk that reform is cyclical.

The optimist's case is genuinely strong: five-year profit compounding near 30% (off the crisis floor, but real), the Citi affluent franchise, a cleaned and granular book, three fat years in a row, and a subsidiary spread across broking, asset management and insurance. The realist's ledger keeps three exhibits pinned beside it: the ₹464 crore year, as a reminder of what this exact institution's leverage did with bad judgment; the 28% dilution, as the price owners paid for it; and the funding-cost gap to the leaders, which compounds against Axis every single day at scale.

Predictability is the sticking point. Buffett pays premiums for banks whose worst year is knowable in advance; Axis's history offers a worst year of ₹464 crore against a best of ₹28,191 crore — a sixty-fold range in a single decade. The franchise is real, national, and improving: narrow moat. The width it lacks is exactly one uneventful boom wide.

Fault lines

Structural risks, in descending order of gravity:

Reversion of credit culture. The bank's own ledger is the base rate. Institutional memory of 2018 is the moat's newest and most perishable course of masonry.
The funding hierarchy hardens. If deposit costs stay stratified by perceived safety, the third-place trust-tax becomes permanent, capping returns below the leaders' regardless of execution.
Systemic embrace. A top-three private bank will carry ever-thicker regulatory capital buffers and obligations; leverage — the entire engine — is precisely what regulation is designed to constrain.
Unsecured retail as the new project finance. The reformed book's growth engine is granular credit; a household over-leverage cycle would correlate those millions of small loans the way one court order once correlated fifty big ones.
Scale's fine print. Contingent liabilities stand at ₹31.2 lakh crore — dominated by the notional value of routine derivatives and trade guarantees, standard plumbing for a bank this size, but plumbing that demands operational excellence forever. Reported profit also includes ₹29,674 crore of other income; owners should track how much of the engine is fees and treasury rather than lending spread.

The amplifier in 2050

Over a generation, India's credit deepening should multiply the banking system severalfold, and the physics of the amplifier will not change: some institution will run nine-to-one leverage over this century's loans, and whoever runs it with boring judgment will compound spectacularly, because the spread times the leverage times the decades is one of the great arithmetic engines available anywhere in commerce. Technology will keep commoditising loan origination while barely touching the two assets that decide banking hierarchies: cheap deposits and institutional temperament.

Axis enters that future as the industry's most complete cautionary tale and its most motivated student — the one large private bank whose own published accounts teach the entire syllabus, from amplification to dilution to repair. If the next boom finds its credit committee saying no — if the worst year of the 2030s is a dull one — the discount this chapter recorded will have marked one of the sector's better entries, and the ledger's strange fourth entry will have been worth every rupee as tuition. The amplifier itself doesn't care either way. It will multiply whatever judgment it is given.

An Omaha Investments chapter. Educational material, not investment advice. Figures from Screener.in and NSE data via Angel One as of the date above.