Tata Consultancy Services: India's biggest programmer toll bridge
If you owned the whole company
Suppose you could write one cheque and buy every share of TCS. The market says that cheque would be about ₹7,44,422 crore. In return, last year the business earned ₹49,454 crore of profit after every bill and every tax was paid.
Do the lemonade-stand arithmetic: pay ₹7,44,422, collect roughly ₹49,454 a year. That is about ₹6.6 back on every ₹100 you put in — a 6.6% "earnings yield." Another way of saying the same thing: the stock trades at a price-to-earnings ratio of 14.2, meaning you pay ₹14.2 today for each ₹1 of current annual profit. And because TCS mails most of its profit back to owners — the dividend yield is 3.07% — you actually receive a good chunk of that in cash every year, not just on paper.
Fourteen times earnings for a business of this quality is not an expensive price. The market clearly has a worry. We will get to the worry.
What does this business actually do?
TCS rents out brains. A big American bank, a European insurer, an airline — their businesses run on enormous piles of computer software: the systems that clear payments, issue policies, book seats. Somebody has to build those systems, fix them when they break at 2 a.m., and slowly modernise them. Most companies don't want to employ fifty thousand programmers themselves. So they hire TCS.
TCS employs an army of engineers, mostly in India, and bills clients — mostly abroad — for the work. It has been doing this for over fifty years, and it is part of the Tata group, which in India means something: clients trust that the firm will still be there, behaving honourably, in twenty years. Last year that trust turned into ₹2,67,021 crore of sales.
The science underneath
There is no chemistry or physics here. Charlie would say the "science" of this business is arithmetic and psychology, and both are worth writing down.
The arithmetic is labour arbitrage. A client in New York would pay a local programmer a very high wage. TCS bills the client noticeably less than that, pays an equally capable engineer in Mumbai or Chennai a fraction of the billed rate, and keeps the spread. Multiply a modest spread per programmer by hundreds of thousands of programmers and you get an operating margin of 27% — for every ₹100 of sales, ₹27 is left after paying the staff and the electricity. That margin has sat between 26% and 28% every single year in our data since 2015. A margin that steady, for that long, at that size, is a machine, not a fluke.
The psychology is switching costs. Here is the part a 14-year-old should memorise: TCS does not just write a client's software — it runs the client's plumbing. When a hundred TCS engineers have spent ten years inside a bank's payment system, they know where every pipe bends. Firing TCS means a new firm must relearn all of it while the water keeps flowing, and if the pipes burst — payments fail, planes don't board — the executive who switched vendors to save 10% gets fired. So nobody switches to save 10%. Fear, not love, keeps the customer. That fear is the moat.
Now the honest third piece: what does AI do to this? TCS sells programmer-hours. AI systems now write serviceable code, which means each engineer can do more per hour — which means clients will eventually ask to buy fewer hours. If your product is the hour, a machine that manufactures cheap hours is aimed at your heart. The hopeful case is that TCS charges for outcomes instead of hours and pockets the productivity itself. The fearful case is that the spread — the whole arbitrage — compresses. We genuinely don't know which way it goes. Neither, we suspect, does TCS. The market's −39% verdict on the stock this past year says it fears the second case.
The moat test
Charlie's test: hand a capable rival ₹7,44,422 crore in cash and ten years of patience. Can they take the castle?
They can hire engineers — engineers are for hire. They can undercut on price. What they cannot buy is fifty years of a client's systems knowledge, the Tata name's trust, or a seat inside a thousand corporations' plumbing. To displace TCS at one bank they must persuade that bank to endure years of risky transition for a modest saving. Multiply that persuasion by a thousand clients. Ten years is not enough; the incumbents' contracts roll on and the knowledge compounds.
Scale helps too: with ₹2,67,021 crore of revenue, TCS spreads training academies, delivery centres, and bench costs across a base no newcomer matches. That is why its return on capital employed is 63% — for every ₹100 of money tied up in the business, ₹63 of operating profit comes out. Castles do not usually earn like that.
Verdict on the moat: wide — against other IT companies. Against a change in what clients buy at all, no moat protects you. Ask the makers of typewriters.
The numbers Warren would check
| What we check | What it means | TCS |
|---|---|---|
| Sales growth, 10 yr | How fast the shop grows | 9% a year |
| Profit growth, 10 yr | How fast owners' earnings grow | 8% a year |
| Return on equity | Profit per ₹100 owners left in | 51.8% |
| Return on capital employed | Profit per ₹100 of all money used | 63.0% |
| Operating margin | Kept from each ₹100 of sales | 26–28%, every year |
| Borrowings | Money owed to lenders | ₹11,283 cr — trivial vs profit |
| Cash from operations | Profit that arrived as actual cash | ₹52,094 cr vs ₹49,454 cr profit |
| Dividend payout | Share of profit mailed to owners | 81% last year |
| Promoter holding | Tata's own stake | 71.77% |
Read that table slowly, because it is close to perfect. Profits arrive as real cash — cash from operations of ₹52,094 crore actually exceeds reported profit, the sign of an honest ledger. Debt is a rounding error. The equity capital line went from ₹191 crore to ₹375 crore in 2018 — that was a bonus issue, free extra shares to existing owners, not dilution — and has since shrunk from ₹366 crore to ₹362 crore as the company bought back its own shares. The owners' slice of the pie is growing, not shrinking.
The one soft number: growth. Screener's own "cons" note says it plainly — sales grew only about 10% a year over five years, and just 6% over the last three. This is a magnificent machine that is no longer young.
What could go wrong
Invert: what kills TCS?
- AI eats the billable hour. Covered above; it is risk number one, and it is real.
- Growth fades to nothing. Ten-year sales growth of 9% has slowed to 6% over three years and 5% in the trailing twelve months. A wide moat around a shrinking castle still houses a shrinking kingdom.
- Visa and politics. Most revenue comes from Western clients; a government hostile to offshore work can tax the arbitrage at the border.
- Wage inflation at home. The spread only exists while Indian engineering salaries stay well below Western billing rates. Every year that gap narrows a little.
- Valuation was the past decade's real lesson. Profit grew 8% a year for ten years, yet the stock returned only 6% a year — buyers a decade ago paid a price that assumed more. The stock is down 39% in a year not because the business broke, but because the price finally argued with the growth.
What management must do to keep the castle
The moat is wide, so this memo is short:
- Sell outcomes, not hours. Get paid for the finished bridge, then build it with AI-assisted crews and keep the savings.
- Keep returning cash — the 81% payout and the buybacks are exactly right for a business that doesn't need the money.
- Protect the Tata trust asset with boring, spotless conduct. It took fifty years to build and one scandal would dent it.
The verdict
Moat: wide. TCS is the finest expression of the Indian IT model: a 51.8% return on equity, margins that haven't wobbled two points in a decade, profits that arrive as cash and get mailed to owners. The switching-cost moat should hold against any rival for five to ten years. The honest caveats are two: the business now grows single-digit slow, and artificial intelligence is a genuine, unresolved threat to the very unit it sells — the programmer-hour. At 14.2 times earnings with a 3% dividend, the price no longer assumes miracles. A wonderful company, finally not priced as one; whether it is still wonderful in 2036 depends on a technology question nobody can yet answer.
Written in the style of Buffett & Munger for the Omaha Investments book project. Educational material, not investment advice. Numbers from Screener.in and live NSE data via Angel One as of the date above.