How an IT Services Business Makes Money (Infosys)

A relative who runs a small labour contracting business once explained his whole model to me over tea, in about a minute. "I have forty masons and carpenters. A builder hires them by the day. I pay each man ₹800, I bill the builder ₹1,100, I keep ₹300. My only job," he said, "is to keep them all working — because an idle mason still eats."

I didn't realise it then, but he had just described how the largest companies on the NSE earn their money. Strip away the glass campuses and the AI press releases, and an IT-services business — Infosys, TCS, Wipro, HCL — is doing exactly what my relative does. It rents out skilled hours at a markup. Understand that one sentence, and the whole sector stops being a mystery.

Let me use Infosys as the worked example throughout. This is a teardown of how the business works, not a view on the share — nothing here is a recommendation to buy or sell it.

The core engine: billing time at a markup

An IT-services firm hires engineers, puts them on a client's project, and bills the client more per hour than it pays the engineer. The gap is the profit. That's it. There's no factory, no inventory, no product on a shelf — the "product" walks in every morning and goes home at night.

You can see the engine in the size of it. In FY26, Infosys earned about ₹1,78,650 crore of revenue and ₹29,474 crore of net profit from a workforce of about 3.29 lakh people (Inve data, 2026; headcount per Infosys's Q4 FY26 results). Divide one by the other: each employee brought in roughly ₹54 lakh of revenue and ₹9 lakh of profit for the year. That is the masons-at-a-markup business at industrial scale — about 3.3 lakh "masons," each billed out for the year.

Because the asset is people, not plant, the business needs very little of its own capital to grow. That is the quietly beautiful part, and it's why Buffett would recognise the shape of it. "The best businesses by far for owners continue to be those that have high returns on capital and that require little incremental investment to grow," he wrote in his 2009 letter to shareholders. An IT-services firm hires more people to grow — it doesn't pour crores into new plants — so it can return most of its profit to owners. Infosys carries almost no debt and routinely hands cash back through big dividends and periodic buybacks (it announced an ₹18,000 crore buyback in FY26, its largest ever, for instance).

Lever one: utilisation — are your masons working?

My relative's whole nightmare was an idle mason. The same fear runs every IT firm. The number that captures it is utilisation — the share of an engineer's available hours that are actually billed to a client. An engineer "on the bench" (hired, paid, but not on a project) is pure cost with no revenue, exactly like a mason eating lunch on a day no builder called.

You won't find utilisation as a line in the financial statements, but you feel its shadow in the operating margin — revenue left after the cost of running the business. Infosys held a reported operating (EBIT) margin of about 21% through FY26 (20.9% in Q4). On Inve's own computed basis it shows nearer 24%, because the two are calculated differently — so always note which margin you're reading. When utilisation slips, that margin slips, because the firm is paying people it isn't billing. When a company says it is "improving utilisation," translate it plainly: fewer people on the bench, more masons working.

Lever two: attrition — and the cost of the revolving door

Now the second fear. If your trained masons keep leaving for the contractor next door, you spend your life hiring and re-training instead of building. In IT this is attrition — the rate at which employees quit each year.

Infosys reported voluntary attrition (IT services) of 12.6% in Q4 FY26, up slightly from 12.3% the prior quarter (Infosys Q4 FY26 results). At 12.6%, roughly one in eight people walks out the door every year and has to be replaced and retrained. During the post-COVID boom, attrition across the sector ran past 20% — wages spiked, margins got squeezed, and every firm was bidding for the same engineers. Low and steady attrition is a good sign; a sudden jump is a warning that costs are about to rise. Watch the direction of the number more than the number itself.

Lever three: deal wins and the order book — next year's work

A contractor with no builders lined up for next month is in trouble, however busy he is today. So the third thing an owner watches is the pipeline of future work — what IT firms call deal wins or TCV (total contract value: the full rupee, or dollar, value of new contracts signed). It's the order book — work booked but not yet delivered.

This is where the recent Infosys record is genuinely strong. On its FY26 results call, management said it signed 96 large deals with total contract value of $15 billion, of which 55% was "net new" — brand-new business, not renewals — and that large deals "were very good, $14.9 bn for the full year… 28% higher than it was in the previous year" (Infosys Q4 FY26 concall, via Inve data, 2026). Net-new deals are the masons you've lined up for builders you've never worked with before. A growing, net-new-heavy order book is the single best forward signal an IT business gives you — far better than this quarter's revenue, which is just yesterday's order book being delivered. One caveat for beginners, though: TCV and deal-win figures are management-reported, and that headline value is spread over several years and is not the same as revenue already booked — so read it as a direction-of-travel signal, not money in the bank.

You can read what management said about deals, AI projects and the pipeline — in plain English, without sitting through the hour-long call — in Inve's concall summaries.

Lever four: the rupee, and why "constant currency" matters

Here is the lever that confuses most beginners. Infosys earns the bulk of its money abroad — Indian IT exports were worth around $233 billion in FY25, the lion's share of a roughly $297 billion industry (IBEF, citing NASSCOM). It bills clients in dollars, euros and pounds, but reports profit in rupees. So two different things move the revenue line: how much work the firm actually did, and what the rupee did against the dollar.

To separate the two, IT firms guide in constant currency (CC) — growth stripped of exchange-rate swings, so you see the real volume of work won, not a currency illusion. When Infosys guides "1.5% to 3.5% revenue growth," it means in constant currency. Always check whether a growth number is CC or rupee-reported; a weak rupee can flatter a firm that actually grew slowly. The business is the work, not the exchange rate.

Putting the levers together — and what the price asks

So an IT-services owner reads four dials at once: utilisation (are my people billed?), attrition (am I bleeding trained people?), deal wins (is next year's work coming in?), and the rupee (is the growth real or a currency mirage?). Margin tells you how well it ran this quarter; the order book tells you about next year; attrition and utilisation tell you whether the engine is healthy underneath.

Then there's price. At a market value of about ₹4.71 lakh crore, Infosys trades at roughly 16 times its FY26 earnings and about 2.6 times its revenue (Inve data, 2026). For a business this asset-light, the market is paying mostly for the durability of the markup — the belief that clients will keep renting these hours, and keep paying a premium for them, for years. The whole bet, in one line, is whether that premium holds as automation and AI change what an "hour of engineering" is even worth.

Where this gets harder

One honest boundary: the levers above tell you how the machine runs, not whether the destination is safe. The harder question for any IT firm right now is demand — will clients keep spending, and will AI tools let one engineer do the work of three (good for margins) or let clients skip the firm entirely (bad for revenue)? Those are judgement calls about the future, not facts on a results sheet, and we won't pretend the numbers settle them.

The discipline an owner can control is narrower and more useful: read management's own guidance, then check the next few quarters to see whether it held. Infosys guided FY26 constant-currency growth of 0–3% and came in at 3.1% — right at the top of the range; the year before, it had guided 4.5–5% and missed (Inve data, 2026). That's not a verdict — two years is a short window — but it's the right habit: take the guidance, mark it on the calendar, and grade the delivery yourself. Tracking that across a whole portfolio, quarter after quarter, by hand, is precisely the chore nobody has time for — which is the job Promise Tracker does.

My relative is still running his crew, by the way. He never reads a balance sheet. But he could explain Infosys to you better than most brokers — because he already knows the only questions that matter. Are my people working? Are they staying? Is next month's job booked? And did the builder pay in the currency I expected? Same business. More zeros.

If the four-dial habit is new to you, two earlier pieces in this series sit underneath it: what you own when you buy a share and return on capital, the quality number.

Frequently asked questions

Related reading

• How a Bank Makes Money: CASA, NIM and Bad Loans
• Order Book: How a Capital-Goods Business Earns
• How a Cement Company Makes Money in India