---
title: Bharti Airtel — Farming the Invisible
symbol: BHARTIARTL
company: Bharti Airtel Ltd.
sector: Telecommunication
moat: wide
date: 2026-07-07
verdict: Physics rations the airwaves into a three-player oligopoly; the survivor now harvests what the war cost it.
---

# Bharti Airtel: Farming the Invisible

## A voice crosses the city

Say "hello" into a phone in south Delhi and follow the word. A microphone
turns the pressure of your breath into a wiggling electrical current. A chip
samples that wiggle eight thousand times a second and converts it into
numbers — your voice is now arithmetic. The numbers are stamped onto a radio
wave, an invisible ripple in the electromagnetic field, the same family of
wave as light, only about a million times longer from crest to crest. The
wave leaves the phone in all directions at the speed of light, weakening as
the square of the distance, and within a kilometre or two it is caught by an
antenna on a rooftop. From there your "hello" abandons radio altogether: it
drops into a glass fibre and travels across the city as pulses of laser
light, emerging at another rooftop, riding one final radio hop into your
friend's phone, where arithmetic becomes wiggle becomes sound. Total elapsed
time: a few hundredths of a second. Total distance travelled by radio: maybe
three kilometres out of thirty. A mobile network is mostly a wired network
wearing a short wireless skirt at each end.

Everything strange and wonderful about the economics of this industry hides
in that last radio hop. So let us stay there a while.

## Land in the sky

A radio wave is described by its frequency — how many times per second it
oscillates. Nature offers a continuous ladder of frequencies, and physics
attaches a cruel trade-off to every rung. Low-frequency waves, with long
lazy crests, bend around buildings and soak through concrete; they carry a
signal deep into a basement, but they oscillate slowly, and a slow
oscillation can only carry so much information — think of signalling to a
friend by waving a flag: the faster you can wave, the more you can say per
minute. High-frequency waves wave furiously and can carry torrents of data,
but they behave more and more like light: they travel in straight lines, die
in the rain, and stop at the first wall.

Now the crucial fact, the one on which this entire chapter balances: **two
transmitters in the same place cannot use the same frequency at the same
time.** Their waves add together in the air and the receiver hears gibberish
— the radio equivalent of two people shouting into one ear. Interference is
not a technology problem awaiting a clever fix; it is a property of waves,
as fundamental as two solid objects refusing to occupy the same space.
Which means the usable radio spectrum over any city is *land*. It has
prime districts (the low bands that penetrate homes), it has far-flung
suburbs (the high bands that carry more but reach less), and there is
absolutely no way to manufacture more of it. Governments understood this and
did what governments do with scarce land: they surveyed it, divided it into
plots, and auctioned it — in India, for sums that have run into lakhs of
crores across the auctions of the past two decades.

Engineers, forbidden more land, learned to farm it more intensively. The
trick that created the modern industry is *reuse*: a frequency exhausted
over one neighbourhood can be used again a few kilometres away, if each
transmitter speaks softly enough. Divide the city into small cells, put a
low-power antenna in each, and the same plot of spectrum yields a harvest in
every cell simultaneously. Want more capacity? Shrink the cells and build
more of them. This is why your phone bill funds a forest of towers: cell
density is the only knob left once the spectrum is bought. And each new
tower needs power, rooftop rent, radio gear, and — remember the wireless
skirt — a fibre running back to the core network. The physics of
interference, followed honestly to its conclusion, decrees that a national
mobile network is one of the most capital-hungry objects a private company
can build.

## Why the market seats three

Walk the logic forward into economics. To serve one customer you must
first buy spectrum for a billion, build towers for every square kilometre
you intend to cover, and lay or lease fibre between all of them. To serve
ten million customers you need — almost the same network. The cost of
carrying one extra call, one extra gigabyte, across infrastructure already
built rounds to zero. Industries shaped like this — colossal fixed costs,
negligible marginal costs — do not stay crowded, because in a crowded
market someone always cuts price toward marginal cost, which is to say
toward nothing, and then everyone bleeds until the weakest die.

The industry's history is the slow discovery of this logic. Marconi
proved in 1901 that a signal could cross the Atlantic without a wire, and
for half a century radio meant broadcasting — one powerful transmitter,
many passive listeners, no scarcity problem worth arguing about. The
scarcity crisis arrived with the desire for *conversations*: thousands of
private two-way calls in one city, each needing its own slice of
spectrum. Engineers at Bell Labs sketched the answer — the cellular
reuse idea described above — as early as 1947, but the electronics to
hand a call from cell to cell as a car drove across town took until the
1980s to build cheaply. India, characteristically, skipped a generation:
having never wired most of its people (the waiting list for a fixed
telephone line once ran to years), it leapt straight to mobile when
licences were issued in the mid-1990s. This company won one of the first
— the Delhi circle — when incoming calls still cost the receiver money
and a minute of talk was priced like a small luxury. Its founder's
lasting innovation was not technical but organisational: outsource the
network hardware and IT to global specialists, keep only the customer,
the brand, and the spectrum, and use the freed capital to buy licences
faster than rivals could. It became the country's largest operator, then
crossed to Africa in 2010 by acquisition, betting the same
low-price, high-volume playbook would work in the world's other great
under-connected market.

Then India ran the marginal-cost experiment at full scale and in
fast-forward. Through the
2000s a dozen operators fought for the world's fastest-growing subscriber
base. Then, in 2016, a new entrant backed by the deepest pockets in Indian
business launched with data priced near zero. What followed was the
industry's marginal-cost logic executed without anaesthesia. Look at this
company's own ledger: revenue of ₹96,101 crore in Mar 2015 *shrank* to
₹80,780 crore by Mar 2019 — four years of a growing country using
exponentially more data while the company selling it earned less. Profit of
₹5,048 crore dwindled to ₹1,688 crore, then vanished entirely: a loss of
₹30,664 crore in Mar 2020 and another ₹12,364 crore in Mar 2021, years
swollen by court-ordered payments of old regulatory dues. A dozen rivals
became, effectively, three — two private survivors and a state-run
operator. Weaker names were absorbed, or simply switched off their
transmitters and handed the spectrum back.

Here is the part a student of business should tattoo somewhere visible:
the brutality and the moat are the same fact. The price war proved that
this market cannot feed more than about three networks — and in doing so
it guaranteed that nobody sane will ever fund a fourth. The barrier to
entry is now a matter of public record, denominated in the losses of the
dead.

## The tower, the fibre, and the bill

What would it take to replicate this business today? Spectrum across
twenty-two licence regions, bought at auction against incumbents who
already earn from theirs. Hundreds of thousands of cell sites, each
negotiated, powered, and maintained. Fibre by the hundreds of thousands of
kilometres. A distribution system that can collect small recharges from
every bazaar in the country. Regulatory standing in a sector where the
rules — dues, auction terms, licence conditions — have historically been
rewritten mid-game. And then, having spent all that, you must persuade
customers to leave networks that already work, in a market where the
incumbents can afford to match any price you offer for longer than your
investors can watch.

The company itself carries the fossil record of this intensity on its
balance sheet. Borrowings climbed from ₹83,415 crore in Mar 2015 to a peak
of ₹2,26,020 crore in Mar 2023 — the accumulated price of spectrum,
towers, and survival — before easing to ₹1,95,412 crore in Mar 2026.
Equity capital rose from ₹1,999 crore to ₹3,047 crore over the decade:
existing owners were diluted, repeatedly, to keep the war chest filled.
Nobody should romanticise this. The moat was paid for with shareholders'
blood, some of it fresh.

## Munger at the auction house

Now reason about the healed market the way a multidisciplinary sceptic
would.

**Incentives, first and always.** A three-player market where every player
carries heavy debt and has already tasted a decade of mutually assured
loss-making has, for the first time in Indian telecom history, a shared
incentive to let prices rise. Tariff increases that would have been
competitive suicide in 2010 now pass through the market like weather —
each operator raises, the others follow within weeks, and customers,
whose monthly bill remains a small slice of household spending for a
service they will not give up, grumble and pay. Observe the operating
margin: 35% in Mar 2015, 57% in Mar 2026. That climb is not a management
miracle; it is the arithmetic of an oligopoly convalescing.

**Switching costs, honestly assessed.** Weak, individually — porting a
number takes days and the rival's SIM costs nothing. The stickiness is
subtler: family plans, bundled broadband, prepaid inertia, and the simple
fact that with only two comparable private networks, the dissatisfied
customer has exactly one place to go — and that place prices like an
oligopolist too. A duopoly does not need loyalty; it only needs the exit
door to lead somewhere equally priced.

**The regulator as silent partner.** Invert the usual question. Ask not
"what makes this business strong?" but "who can make it weak overnight?"
The answer is not a competitor. It is the state — which sells the spectrum,
sets the dues, taxes the sector, owns the fourth operator, and once
presented a retrospective bill large enough to produce that ₹30,664 crore
loss. The government has since behaved like a shareholder who wants the
industry alive, converting some dues to equity in the weakest player and
spacing out payments. But a moat whose gatekeeper is sovereign is rented,
not owned.

**The decade's to-do list, were Munger chairing the board:** keep cutting
debt while the tariff sun shines — the fall from the 2023 peak is the
right direction and must continue; convert the fibre-and-tower estate into
recurring enterprise and home-broadband revenue, which rides the same
physics with stickier customers; make the Africa operations — the company
serves seventeen countries beyond India — compound quietly rather than
demand capital; resist the temptation, when the balance sheet finally
breathes, to diversify into glamour; and treat every spectrum auction as a
discipline test, buying what the physics requires and nothing for pride.

## What the survivor earns

Look at the business now through an owner's eyes, the whole company at
once. The market values it at ₹11,72,987 crore — about 40.8 times the
last year's earnings and 7.84 times book value. For that price you receive
a machine that turned ₹2,10,973 crore of revenue into ₹33,823 crore of
profit in Mar 2026, converts earnings into cash with unusual force —
operating cash flow has quadrupled in six years, ₹28,059 crore in Mar 2015
to ₹1,22,230 crore in Mar 2026, because customers prepay and the network's
costs are largely fixed — and earns a return on equity of 21.9% against a
10-year average of 12%, that gap being the whole story of before-and-after
the war. Five-year profit growth of 24% a year rests substantially on
tariff repair, a well that is deep but not bottomless.

The honest owner also counts the scars. A decade of sales growth averaging
just 8% a year embeds those four shrinking years. Dividends have been
erratic — the payout ratio has swung from 244% to zero to 55% — because
cash always had a prior claimant: the network, the auction, the lender.
Promoter holding has drifted from 54.75% to 50.07% between Sep 2023 and
Jun 2026. And the debt, though falling, still stands at nearly twice the
company's reserves of ₹1,46,010 crore.

Is it a business that can compound for decades? The demand side is as
predictable as anything in this book: Indians will use more data every
year for the rest of our lives, and the physics of interference ensures
they can buy it from only a few sellers. The rating here is **wide** —
a moat made of auctioned spectrum, sunk towers, and the documented
corpses of everyone who tried to make it a competitive market. The width
is real; the caveat is that one of the walls belongs to the government.

## Three ways to lose the airwaves

The structural threats deserve prose, not a list, because they intertwine.
The first is the state itself, already discussed: retrospective dues,
auction design, or a policy tilt toward the operator it part-owns could
tax away the oligopoly's peace. The second is a change in the physics of
access: satellite constellations that beam broadband directly to ordinary
phones are no longer science fiction, and while beaming from five hundred
kilometres up cannot match the capacity of a tower five hundred metres
away — the interference arithmetic favours density — satellites could
skim the rural coverage premium and cap pricing power at the edges. The
third is self-inflicted: capital cycles in telecom have always rhymed, and
the same executives who swear discipline today will feel the old itch when
the next generation of technology is auctioned. The moat's worst enemy has
historically been the industry's own chequebook.

## The invisible estate in 2050

Two structural forces run in this company's favour for decades. First,
traffic: every trend of the coming thirty years — video, AI assistants
that see and listen, machines talking to machines, a billion Indians
climbing the income ladder — is a traffic trend, and all traffic must
cross somebody's spectrum and fibre. Second, consolidation of everything
digital *onto* the network: payments, entertainment, enterprise computing.
The network operator does not need to win those businesses to win from
them; it is the ground they are built on.

The technology will keep churning — generations of radio standards will
come with new letters and new auctions — but notice what never changes
beneath the churn: the electromagnetic spectrum over India is the same
size in 2050 as it was when Marconi first sparked a signal across the
Atlantic. More people, more machines, more data, forever — pressing on a
resource fixed by the laws of physics and leased by the state to a handful
of tenants. Farming the invisible turns out to be the oldest business
model there is: own scarce land near where the people are, and charge
rent. The rent was unpayable during the war. The war is over, and the
land remembers who held it.

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