- TCS, Infosys and Wipro spent roughly 1.09 lakh crore rupees on buybacks since 2020, against combined FY25 R&D of about 4,350 crore, around twenty-five rupees returned for every one invested in R&D.
- The three majors spend under 1% of revenue on R&D: TCS about 1%, Infosys under 1%, Wipro closer to half a percent.
- Global AI builders invest far more: Meta puts about 27% of revenue into R&D, Alphabet about 14%, Microsoft about 12%, and Meta's single-year R&D budget tops the Indian trio's six-year buyback bill more than threefold.
- With AI eroding the labour-arbitrage model and growth slowing, the market has started to reprice Indian IT and FIIs have trimmed exposure.
- The core issue is capital allocation: buybacks chosen over building the future.
Two numbers tell you most of what you need to know about Indian IT right now.
Since 2020, the country’s three largest IT firms, TCS, Infosys and Wipro, have spent roughly ₹1.09 lakh crore buying back their own shares. Their combined research-and-development spend in FY25 was about ₹4,350 crore. For every rupee these companies put into building the future in a year, they have handed back something like twenty-five over the buyback cycle.
That is not a rounding error. That is a strategy. And the market has finally started to price it.
The contrast that matters: R&D versus buybacks
Reduce it to a single ratio, R&D as a share of revenue, and the gap becomes very hard to defend. TCS spends about 1% of revenue on R&D. Infosys, under 1%. Wipro, closer to half a percent.
Now look at the companies actually defining the AI era. Meta puts roughly a quarter of its revenue into R&D, closer to 28% in recent quarters. Alphabet runs at about 14%. Microsoft, around 12%. Meta’s R&D budget in a single year, near $44 billion, or about ₹3.7 lakh crore, is larger than the entire six-year buyback bill of India’s three biggest IT firms. By more than three times over.
| Company | R&D (latest FY) | % of revenue |
|---|---|---|
| Indian IT majors (FY25) | ||
| TCS | ₹2,630 cr | ~1.0% |
| Infosys | ₹1,296 cr | ~0.8% |
| Wipro | ₹431 cr | ~0.5% |
| Global AI builders (latest FY) | ||
| Meta | ~$44 bn | ~27% |
| Alphabet | ~$49 bn | ~14% |
| Microsoft | ~$30 bn | ~12% |
R&D as reported in company filings, Indian IT for FY25, the US firms for their latest fiscal year. The comparison metric is R&D as a share of revenue.
There is no version of an AI strategy that starts with buying your own stock.
This is not one bad quarter
It is tempting to read the weakness in IT as a demand pause: clients deferring discretionary projects, a soft patch that mean-reverts in a quarter or two. That was a fair reading in 2023. It is much harder to argue now, two years on.
The sector trades in the high teens on earnings, below its long-run average, but not cheap for a business growing revenue in low single digits in constant-currency terms. IT’s weight in the Nifty 50 has slipped to multi-year lows, down from nearly a fifth of the index at the 2021–22 peak. Foreign investors are rotating toward Korea, Taiwan, Japan and the US for cleaner AI exposure. This is not the market reacting to one print. It is the market repricing a model.
Five fault lines
The pressure is not coming from one direction. Five things are giving way at once.
The revenue model
Indian IT was built on labour arbitrage, billable hours at a cost the West could not match. AI compresses exactly that layer. Bill rates are softening before the deal wins even reflect it. When the unit of value is “hours,” and software is starting to do the hours, the whole pricing structure is exposed.
Demand
Discretionary spending across BFSI, retail and healthcare has stayed frozen far longer than the “one-to-two-quarter pause” everyone promised. Worse, the budgets that are moving are increasingly written straight to OpenAI, Anthropic, Microsoft and Google. The client’s AI line item is no longer routed through a services vendor.
The competition it created
India now hosts close to 1,900 global capability centres employing around two million people. Global banks and big tech are building their own AI and engineering hubs in Bengaluru and Hyderabad, hiring from the same talent pool, for the same work. Indian IT’s clients are quietly becoming Indian IT’s competitors.
The innovation gap
Topaz, AI Studio, the various “platforms”, incremental, not category-defining. Still services, still no owned IP, still no products of consequence. Western consulting has been buying AI capability aggressively; Indian IT has been slower, and is trying to build the future on under 1% of revenue.
People and politics
A headcount-led model breaks the moment AI cuts the hours per project, bench utilisation suffers, promised hikes squeeze margins, and the attrition story flips from a hiring problem to a layoff problem. On top of that, the US H-1B petition fee has been pushed to around $100,000, Indo-US services trade is uncertain, and EU AI rules add compliance overhead to an already thin model.
Capital allocation is the tell
Here is the part the buyback numbers give away.
| Company | Buyback programmes since 2020 | Total |
|---|---|---|
| TCS | 2020, 2022, 2023 | ₹51,000 cr |
| Infosys | 2021, 2022, 2025 | ₹36,500 cr |
| Wipro | 2020–21, 2023 | ₹21,500 cr |
| Combined, since 2020 | ₹1,09,000 cr | |
Cumulative announced buyback value since 2020. Against this, the three firms’ combined FY25 R&D was about ₹4,350 crore.
When a management team returns cash year after year, TCS alone has run multiple programmes in the ₹16,000–18,000 crore range, it is telling you something. It is saying it cannot find a better use for that capital than buying its own stock. In an ordinary year, that is shareholder-friendly discipline. In the middle of an AI capex supercycle, when every global peer is spending double-digit percentages of revenue to build the next platform, it is closer to an admission.
The cash has been going out, not into the future. That is the signal. Not the guidance, not the commentary on the earnings call, the cheque.
What is holding up the multiple
The bull case has narrowed to two lines.
The first is the rupee cushion. A weaker rupee flatters reported revenue and margins, but that is a reporting effect, not a fix. It does not create a single new dollar of demand or a single unit of owned IP.
The second is balance-sheet quality and cash generation, and that part is genuinely real. These are clean, cash-rich businesses. But a strong balance sheet does not justify a premium multiple when growth is missing, the revenue model is being openly questioned, and the surplus is being handed back rather than reinvested.
The view on Indian IT
The market is not punishing Indian IT for a single quarter. It is repricing a sector that chose to return its cash while the rest of the world was busy building. Until that capital-allocation choice changes, until R&D as a share of revenue starts to look like a company that intends to compete in AI rather than harvest the last of the arbitrage, rallies in Indian IT stocks are more likely to be sold into than bought.
The balance sheets are strong. The strategy is the problem. And no buyback has ever bought a company its way into the future.
