---
title: Tech Mahindra — Scar Tissue, Signal Towers, and the Machine That Types
symbol: TECHM
company: Tech Mahindra Ltd.
sector: Information Technology
moat: narrow
date: 2026-07-07
verdict: Sticky contracts, telecom concentration, and visible scar tissue; the recovering patient in a family of athletes.
---

# Tech Mahindra: Scar Tissue, Signal Towers, and the Machine That Types

## The letter that arrived on a January morning

On the morning of 7 January 2009, a letter reached the stock exchanges
of India. Its author was B. Ramalinga Raju, chairman of Satyam Computer
Services — then the country's fourth-largest IT firm, auditor-approved,
award-winning, employed by a third of the Fortune 500. The letter said,
in effect: the cash on our balance sheet does not exist. Over a billion
dollars of it was invented. The confession's most quoted line compared
the fraud to "riding a tiger, not knowing how to get off without being
eaten."

Indian IT's entire product — remember the chapters before this one — is
trust sold to strangers eight time zones away. That morning, the product
itself was in doubt. The government seized the company, auctioned it to
save fifty thousand jobs and the industry's name, and the winning bidder
was a mid-sized telecom-software specialist from the Mahindra group. The
minnow swallowed the whale, spent years fusing the two, and the merged
firm — Tech Mahindra — is the company in this chapter.

That origin story explains almost everything structural about Tech
Mahindra: why it is the most concentrated of the majors, why its
earnings carry visible scar tissue, and why it makes the ideal specimen
for this book's hardest question — the one we have deferred four
chapters and will now face head on: what happens to a business that
sells programmer-hours when machines learn to program?

## A network is a promise

First, the science of the company's home turf, because Tech Mahindra
was born a telecom specialist and remains defined by it.

A telephone network is engineering's most extreme promise: that any of
billions of devices can reach any other, within seconds, essentially
always. Keeping that promise is mostly invisible software. When your
phone rings, computers have already agreed — in milliseconds — which
towers, switches, and cables will carry the call, what to do if a link
fails mid-sentence, and how to hand your moving car from tower to tower
without a click. Telecom software is written to a standard most
industries never attempt: "five nines" availability — 99.999% uptime,
about five minutes of failure allowed per year — because a network that
drops emergency calls is not a product with a bug, it is a public
hazard.

Software this critical, this specialised, and this perpetually
upgraded — every decade a new generation: 2G, 3G, 4G, 5G, each requiring
the network's brain to be rebuilt while it keeps running — created a
natural outsourcing market. Telephone companies are superb at spectrum,
towers, and billing; they have never much wanted to employ armies of
signalling-software engineers. In 1986, the Mahindra group and British
Telecom formed a joint venture to supply exactly that army from India.
For its first two decades the company was, in essence, BT's engineering
wing at Indian prices — a fine business with one enormous embedded flaw,
which economics students will recognise instantly.

## The arithmetic of one big customer

The flaw is concentration, and it deserves its own movement because Tech
Mahindra's financial history is a controlled experiment in it.

A services firm with one dominant client enjoys wonderful economics —
deep knowledge, no sales costs, renewals by habit — and has surrendered
the one thing that makes economics wonderful durably: bargaining power.
The client knows your dependence to the decimal. Worse, you inherit the
client's industry cycle. Telecom operators live a brutal treadmill:
every decade they must buy spectrum and rebuild networks at colossal
cost, while competition holds tariffs down; when their returns sag,
they squeeze suppliers — and the engineering-services line is among the
easiest to squeeze.

The Satyam acquisition in 2009 was, among other things, the cure: it
brought manufacturing, banking, and healthcare clients that diluted
telecom from the overwhelming majority of revenue to something like
a third — still, decades later, roughly double the telecom exposure of
any peer. So when the telecom cycle turned harsh and the company's own
execution wobbled, the ledger recorded it without mercy. Net profit:
₹5,630 crore in Mar 2022, then ₹4,857 crore, then ₹2,397 crore in Mar
2024 — a fall of more than half in two years, to below the ₹2,659 crore
the company had earned in Mar 2015. Nine years of growth, round-tripped.
Operating margin tells the same story in one line: 18% in Mar 2022,
9% in Mar 2024 — and TCS, in the same two years, held 26–27% without a
tremor. That is what concentration plus average execution does when the
cycle turns. The repair since is real — margin back to 16%, profit to
₹4,806 crore in Mar 2026 — and the scar is the honest teacher here:
diversification you postpone is charged for later, with interest.

## The machine that types

Now the deferred question. Across this book's IT chapters we have
described an industry whose product is trust and whose unit of sale is
the engineer-hour. In the last few years, a new kind of software —
large language models, machines trained on essentially everything
humans have written, including all the world's public code — has
learned to produce working programs from plain-English requests. Let us
think about it the way Feynman would: no hype, no dread, just
mechanics.

What does such a machine actually do? It converts a *precise
description* of a program into the program, almost instantly, almost
free. What does it not do? It does not know what the bank actually
needs, which regulations apply, which of the client's seventeen
systems the new code must not disturb, or who answers when the output
is confidently wrong — and these machines are, by construction, capable
of being confidently wrong. In the industry's cost structure, the
machine attacks the *typing* — historically the most billable bulk of
the pyramid — and leaves the *specifying, verifying, integrating, and
guaranteeing* untouched, or even enlarged.

So the structural question is not "does the work disappear?" It is:
*who captures the productivity?* When a task that took a hundred
billed hours takes twenty, either the vendor keeps billing hours and
shrinks, or it reprices to outcomes and pockets some of the surplus, or
competition passes the whole surplus to clients. History's precedent —
every prior automation wave in this industry, from code libraries to
offshore tooling — says the surplus mostly went to clients, and vendors
survived on volume growth. The AI wave is larger and faster, and this
time volume growth in Western technology budgets is no longer
guaranteed. For a firm like Tech Mahindra — subscale beside the giants,
margin-thin, telecom-heavy — the honest statement is: AI is a bigger
threat to the weak than to the strong, because repricing an industry
punishes whoever has the least room for error. The same honesty
requires the other half: telecom networks are becoming software through
and through (a 5G network is closer to a data centre than to a
switchboard), and machines that write code make network-software
projects cheaper to attempt — which expands the market Tech Mahindra
knows best. Both statements are true. The decade will decide their
relative sizes.

## Munger's reading of the wounds

Munger claimed the most useful records are the painful ones. Read Tech
Mahindra's dispassionately.

Return on equity: 17.5% last year, but 13% averaged over three years —
the dip years drag it — against 24% at HCL, 31.9% at Infosys, 51.8% at
TCS. Ten-year profit growth: 5% a year. Earnings per share: ₹57.27 in
Mar 2022, ₹24.14 in Mar 2024, ₹49.10 now — a round trip that took four
years of shareholders' time. This is what a narrow moat with average
execution looks like when photographed over a full cycle: the clients
stayed (switching costs are the industry's shared inheritance, and Tech
Mahindra's five-nines telecom relationships are genuinely hard to
displace), yet staying power alone put no floor under profitability.

Incentives and ownership: the promoter, Mahindra & Mahindra's group,
holds 34.97% — enough for control and long horizons, small enough that
the company has repeatedly recruited outside professional leadership,
including, in recent years, expressly to run a turnaround. One
incentive detail is telling: in the worst year, Mar 2024, the company
paid out 150% of its profit as dividends — more than it earned — and
94% in each of the two years since. Read generously: a signal of
confidence and a discipline on management. Read sceptically: a company
distributing like a mature utility while needing to invest like a
challenger. Both readings fit the facts; an owner should hold them
simultaneously.

What must management do this decade? First, finish the margin repair
and prove it is structural: the road from 9% to 16% is behind; the road
from 16% toward the leaders' altitude requires pricing discipline and
delivery quality sustained for years, not quarters — the single
clearest test of this franchise. Second, convert the telecom scar into
the telecom option: as networks become software and AI floods into
network operations, be the indispensable engineering partner for that
rebuild — it is the one arena where Tech Mahindra's specialist depth
outranks the giants'. Third, keep diversifying revenue until no client
industry can halve the company's profit again. Fourth, treat AI-led
delivery as existential, not promotional: a subscale firm's only route
through an industry repricing is to be *earlier* than the leaders, not
merely present. Fifth, keep the balance sheet boring — at ₹2,186 crore
of borrowings against ₹29,173 crore of reserves, it is; turnarounds
fail when leverage joins them.

## An owner counts the scars — and the price

The Buffett exercise, with a twist unique in this book's IT quintet.
The whole company sells for ₹1,37,698 crore — 27.5 times last year's
₹4,806 crore of profit. Pause on that: the *weakest* ten-year record of
the five majors carries the *highest* multiple, nearly double TCS's
14.2. The market is not grading history; it is pricing the recovery —
paying today for margins it expects tomorrow. Whatever one thinks of
that bet, the owner should recognise what is being bought: not a
proven compounding machine at a fair price, but an improvement story at
a full one. This book does not forecast; it notes that Buffett built a
career on the observation that turnarounds seldom turn as scheduled.

What the owner verifiably holds: ₹56,815 crore of revenue with genuine
stickiness; operating cash flow of ₹6,172 crore, comfortably above
profit; near-zero debt; a 3.63% dividend yield; and a demonstrated
institutional will to distribute. Domestic institutions have been
buying conviction in the repair — their stake rose from 26.86% to
37.34% in under three years while foreign institutions and 1.9 lakh
retail shareholders departed. What the owner does not hold: pricing
power, scale parity, or a decade of evidence that this firm earns more
than a modest premium over its cost of capital across a full cycle.

Verdict: **narrow moat** — real switching costs and rare telecom-
software depth, guarded by the thinnest execution record among the
majors; the moat kept the clients, and it alone could not keep the
profits.

## What stays broken if nothing changes

The structural risks, in Tech Mahindra's specific order:

- **Industry repricing meets subscale margins.** The AI transition
  will compress the billed hour across the industry; the firm entering
  that compression at 16% operating margin has half the cushion of the
  leader at 27%. This is the compounding of two risks into one.
- **Concentration, still.** Telecom remains an outsized share of
  revenue, and telecom operators remain structurally poor customers —
  capital-hungry, tariff-capped, and ruthless with suppliers in every
  downturn. The Mar 2024 ledger is the rehearsal, on file.
- **The talent pyramid inverts.** All Indian IT recruits huge junior
  cohorts and prices their hours out at a spread. AI consumes junior
  work first, breaking the economics of the base of the pyramid — an
  industry-wide risk that bites hardest where margins are thinnest.
- **The common climate.** Western client dependence, visa and data-
  localisation politics, Indian wage convergence — the whole industry's
  shared weather, in full.
- **Distribution versus reinvestment.** Payouts above earnings during
  a rebuild are a promise that the rebuild is cheap. If the AI decade
  proves expensive, something yields — and dividends, once promised,
  are politically the last to.

## Signal towers, decades out

Look twenty or thirty years ahead and the company's two defining facts
point, unusually, in the same direction. Fact one: the world's networks
are becoming pure software — 5G and whatever follows are data centres
with antennas, rebuilt continuously, and the engineering of that
perpetual rebuild is precisely the craft this company has practised
since 1986. Fact two: machines now write code — which makes every
software-intensive rebuild cheaper to attempt, multiplies the number of
systems in the world, and shifts the scarce skill from typing programs
to specifying, integrating, and guaranteeing them inside brutally
unforgiving environments. Five-nines environments are where "confidently
wrong" is least tolerable and accountable engineering is priced
highest.

There is, in other words, a future in which Tech Mahindra's narrow,
scarred specialisation turns out to be the right shape for the age of
machine-written software. There is also a future in which the same
machines and the same giants squeeze a subscale generalist into
permanent mediocrity. The company's own history — a joint venture that
outgrew its parent client, a minnow that swallowed a fraudulent whale
and made it honest — argues it survives passages. Its ledger argues
survival is not the same as compounding. Both arguments are before the
court, and the decade now beginning is the evidence.

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