---
title: Tata Consultancy Services — The Business of Being Impossible to Fire
symbol: TCS
company: Tata Consultancy Services Ltd.
sector: Information Technology
moat: wide
date: 2026-07-07
verdict: India's finest business machine — half the capital of rivals, triple the return; growth, not quality, is the question.
---

# Tata Consultancy Services: The Business of Being Impossible to Fire

## The bank that cannot leave

Start with a puzzle. Somewhere in Mumbai — or London, or New York — sits a
large bank. Every year it pays Tata Consultancy Services hundreds of crores
to write, run, and repair its software. Every year, a rival firm offers to
do the same work for twenty percent less. The bank's procurement department
exists precisely to accept offers like that. It almost never does.

Why not? The bank is not sentimental. The answer is the key to this entire
chapter, and it is worth stating plainly before we earn it properly: the
bank is not buying code. Code is cheap. The bank is buying the *continued
correct behaviour of a system it no longer fully understands* — and TCS is,
in many cases, the only organisation on Earth that understands it. Firing
TCS would be like firing the only surgeon who has read your medical file,
mid-operation, to save money on the anaesthetist.

To see why such a strange dependency arises — why it is normal, not a
scandal — we have to look at what software actually is.

## Instructions that never forgive

A computer is the most obedient machine ever built, and the least
forgiving. It does exactly what its instructions say — not what the author
meant, not what a reasonable person would infer, but what the text says. A
program is simply a very long list of such instructions. A bank's core
system — the software that knows your balance, moves your salary, computes
your interest — can run to tens of millions of lines, written by thousands
of people over decades, most of whom have retired.

Here is the property that creates the industry. Physical machines wear
out; software does not. A program that worked yesterday will work
identically tomorrow — *if nothing around it changes*. But everything
around it changes constantly: regulations, tax rules, other systems it
talks to, the hardware beneath it, the fraudsters probing it. Software
does not rust; it is rusted *upon* by a moving world. So every large
organisation faces a permanent, growing task: keep the instructions
correct while the world redefines "correct" every month. Computer
scientists call the ongoing cost of past decisions *technical debt* — and
like all debt, it accrues interest whether or not you acknowledge it.

Now add the second property: in a bank, an error is not an inconvenience.
An instruction that rounds interest wrong in the eighth decimal place,
executed a billion times, is a heist. The people who maintain such systems
must combine the caution of accountants with the craft of engineers, in
teams large enough to cover every subsystem, around the clock, for
decades. No bank wants to employ fifty thousand such people directly. It
wants to *rent* them — from someone who trains them, deploys them,
replaces them when they leave, and accepts contractual blame when things
go wrong.

That rental business is TCS.

## Selling certainty by the hour

Notice what is being sold. Software itself has a famous economic property:
once written, copying it costs nothing. That is why a pure software
product company — a Microsoft — can earn margins that look like misprints.
But TCS mostly does not sell copies. It sells hours of skilled attention
applied to one client's unique tangle of systems, and an hour, unlike a
copy, cannot be duplicated for free. Every additional rupee of revenue
requires additional trained people. This is the quiet truth of the Indian
IT industry: it is a people business wearing a technology costume, and its
costs rise in a straight line with its sales.

That truth sets a ceiling on margins — no services firm will ever earn
like a software product firm — but it also builds the floor under the
business. Because the value delivered is not the code but the accumulated,
undocumented knowledge of *this client's* systems, the supplier who has
served a client for fifteen years holds an asset no competitor can buy at
any price: memory. The economics of the industry are therefore a strange
inversion of manufacturing. In steel, the barrier is the cost of building
the plant. Here, the barrier is the cost — and terror — of *changing the
operator* of a plant made of logic, invisible, and load-bearing.

## Sixty years before it was obvious

TCS was founded in 1968, inside the Tata group, at a time when the idea of
India exporting software would have sounded like a joke with no punchline.
India had almost no computers. What it had was mathematics teaching,
English, and engineers priced far below their Western peers. Through the
1970s and 1980s the company learned its trade the humble way — sending
programmers abroad to work on clients' machines, then slowly persuading
clients that the work could travel to India while the responsibility
stayed contractual.

Two outside shocks then built the industry's runway. The first was the
Y2K remediation wave of the late 1990s, which forced thousands of Western
firms to let Indian engineers into their oldest, most sensitive code —
and, crucially, to discover that the work came back correct. The second
was the telecom collapse in prices of intercontinental bandwidth, which
made a programmer in Chennai functionally adjacent to a data centre in
Ohio. Trust plus bandwidth created the offshore delivery model, and TCS —
oldest, largest, and wearing the Tata name — became its flagship. The
group still owns the ship: promoters hold 71.77% of the company today.

It is worth understanding what the "delivery model" actually is, because
it is the industry's one genuine invention — a piece of organisational
technology, not computing technology. Software work is broken into
pieces by risk: the client-facing sliver stays onshore, near the buyer;
the bulk moves offshore, where cost is lowest; and the whole is bound
together by process discipline borrowed from manufacturing — documented
procedures, quality gates, measured defect rates — so that ten thousand
engineers who have never met the client can work on the client's most
sensitive systems without ever dropping the glass. Because India's
working day sits opposite America's, work follows the sun: a problem
logged in New York at closing time is often solved in Mumbai before New
York wakes. None of this is patentable. All of it took decades to
industrialise, and TCS industrialised it first and at the largest scale.

## A fortress with no walls to photograph

What would it cost to replicate TCS? The strange answer: almost nothing,
and everything. There is no refinery to build. A rival needs offices,
laptops, and people — capital a mid-sized fund could supply. Indeed
thousands of small IT firms exist. What cannot be bought is the position:
decades-long relationships inside hundreds of the world's largest
companies; a reputation for never dropping the glass; the machinery to
recruit, train, and organise engineers by the hundred thousand; and a
brand — Tata — that a nervous CIO can say aloud in a boardroom as a
complete answer to the question "and what if this goes wrong?"

The balance sheet tells you how little physical capital the fortress
needs. Against ₹2,67,021 crore of annual sales, the company carries
borrowings of just ₹11,283 crore — a rounding error beside reserves of
₹1,06,878 crore. The moat, such as it is, is made entirely of memory,
reputation, and organisational muscle. That is worth pausing on, because
moats made of intangibles have a special property: they cannot be
destroyed by a competitor's chequebook, only by the owner's own neglect —
or by a change in the underlying technology. Hold that second clause; we
will need it.

## The psychology of the person who signs

Now let us reason as Charlie Munger would — through incentives first,
because incentives explain the bank in our opening puzzle better than any
technology does.

Consider the chief information officer deciding whether to replace TCS
with a cheaper vendor. If the switch succeeds, the bank saves perhaps a
fifth of one budget line, and the CIO gets a nod at the annual review. If
the switch fails — if, during the two-year handover, payments stall for a
day — the CIO's career ends and the bank makes the newspapers. The payoff
is a small, deferred gain against a small chance of professional death.
No rational executive takes that bet to please procurement. Multiply this
person by ten thousand companies and you have the industry's true moat:
*switching costs enforced not by contract but by fear, and the fear is
reasonable.*

Second discipline: scale economics, which in this industry appear in an
unusual place. TCS's size lets it maintain deep teams in every industry,
every technology generation, every time zone — so it is always a safe
choice, which wins it the next large contract, which funds the depth. The
feedback loop compounds quietly. Its financial signature is visible in
the returns: 51.8% return on equity — profit earned on each rupee owners
left in the business — and 63.0% return on capital employed. Numbers like
these mean the business needs almost no retained capital to grow, so
nearly everything it earns can be handed back. And it is: dividend payout
has run between 58% and 100% of profit in recent years, and the share
count has been shrinking through buybacks — profit grew 2.5 times over
eleven years, but earnings per share grew 2.7 times, from ₹50.68 to
₹136.01, because the pie was cut into fewer slices as it grew.

Third: culture and succession. TCS is that rarity, an institution rather
than a founder's shadow — leadership has changed hands repeatedly without
drama, under a promoter whose stated horizon is generational.

What must management do this decade? Three things, honestly ranked.
First, solve the growth problem, because the compounding arithmetic has
slowed: sales grew 9% a year over ten years but only 6% over three, and a
9%-growth business, however profitable, is a bond with a good conscience.
Second, convert the AI threat into the next maintenance annuity — position
TCS as the firm that deploys, supervises, and takes liability for
machine-written code, so that the client's dependency transfers to the new
technology rather than dissolving with the old. Third, resist the
conglomerate temptation to buy growth with the cash pile; this business's
beauty is precisely that it never needed acquisitions to compound.

## What the owner of the whole thing would see

Buffett's exercise: buy every share. The market asks about ₹7,44,422
crore — 14.2 times last year's profit of ₹49,454 crore, a starting
earnings yield of about 7%, with a 3.07% dividend yield paid to you in
cash while you wait.

What do you own? A machine of extraordinary consistency. Operating margin
has stayed inside the 26–28% band for every one of the last twelve years
— through a pandemic, through wage inflation, through technology fashion
cycles. Profit has risen in ten of the last eleven years, from ₹20,060
crore to ₹49,454 crore. Cash conversion is the kind auditors dream about:
₹52,094 crore came in from operations last year, *more* than reported
profit — customers pay, promptly, in cash. There is no inventory, no
factory to rebuild, no commodity input to hedge. Return on equity above
50% means growth is self-funding many times over.

The share register adds its own footnotes. Equity capital nearly
doubled in Mar 2019 — from ₹191 crore to ₹375 crore — through a bonus
issue, which cuts the same pie into more slices for the same owners and
changes nothing real; since then the capital has shrunk steadily to
₹362 crore as buybacks retired shares for cash, which changes something
very real: each remaining slice grows. Ownership has been migrating
homeward — foreign institutions trimmed from 12.46% to 9.66% over the
last three years while domestic institutions rose from 9.80% to 13.34%
— with some 24.5 lakh individual shareholders along for the ride. And
the anchor never moves: Tata Sons' stake makes TCS the engine room of
India's oldest conglomerate, the dividend stream that funds much of the
group — a fact which aligns the controlling owner's deepest interest
with exactly what minority owners want: decades of reliable cash.

The honest owner also records what the machine is not. It is not fast:
8% profit growth over a decade is respectable, not thrilling. And its
revenue line is rented, not owned — every year the hours must be sold
again, priced against hungry rivals, in a currency (skilled labour) whose
cost only rises. The market has noticed the slowing: the share price
compounded at just 6% over ten years and fell hard in the last one. This
book does not do share prices; it does note that a 14 multiple on a 50%-
return-on-equity business is the market saying, in effect, "wonderful
machine, doubtful future." Whether that doubt is right is the risk
section's job.

The verdict on quality, though, is not close. By return on capital, cash
generation, balance-sheet cleanliness, and consistency, this is arguably
the finest large business in India — a **wide moat**, built from switching
costs and trust rather than steel.

## The one honest threat

Structural risks, in order of seriousness:

- **Machines that write code.** For the first time since 1968, the input
  TCS resells — skilled programming hours — has a plausible substitute.
  Artificial intelligence systems now produce competent code from plain-
  English instructions. Stated coldly: a business that bills by the hour
  faces a technology whose purpose is to reduce the hours. The bear case
  is that clients demand the productivity as price cuts, and revenue
  deflates even as work grows. The bull case is the lesson of this whole
  chapter: clients never paid for code; they paid for *accountability*,
  and machine-written code needs supervision, integration, and someone to
  sue — services all. Both cases are partly true. What is certain is that
  the pricing unit of the industry — the hour — will not survive the
  decade unchanged, and repricing an industry is dangerous even for its
  strongest firm.
- **Concentration of destiny.** The clients are overwhelmingly Western
  corporations. A prolonged Western downturn, or political hostility to
  offshoring — visa regimes, data-localisation laws, "hire local"
  mandates — squeezes the whole model from outside.
- **The wage-gap slowly closing.** The original arbitrage — Indian
  engineering cost against Western billing rates — narrows every year as
  Indian wages rise. Scale and switching costs replaced it as the moat,
  but the tailwind of the first forty years is largely spent.
- **Institutional complacency.** Intangible moats die of neglect. A
  decade of harvesting margins while under-investing in the next
  technology wave is how empires of trust end — slowly, then suddenly.

## The long game from here

Look out twenty or thirty years. The world's stock of software will be
vastly larger — every company is becoming, in effect, a software company
that does something else on the side — and all of it will need what
banks' systems have always needed: someone accountable for its continued
correct behaviour. The nature of the work will shift, perhaps radically:
fewer humans typing instructions, more humans (and machines) verifying,
integrating, and governing what machines produce. The industry's revenue
may deflate per task even as tasks multiply.

TCS enters that transition with the strongest hand at the table: the
deepest client memory, the cleanest balance sheet, a brand that means
"nothing will go wrong," and a return on capital that funds any
reinvention out of petty cash. The company's history is one of surviving
its own product becoming obsolete — punch cards, mainframes, Y2K, the
web — by selling, each time, the same underlying thing: certainty, by
whatever unit certainty is currently measured in. The bank in our puzzle
still cannot leave. The question the next decades will answer is whether
the thing it cannot leave remains expensive.

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