The IT crash: GenAI break, or sentiment overshoot?

In 2026, the Nifty IT index fell to a 30-month low — down roughly 25% for the year. TCS shed nearly ₹70,000 crore in market value. Infosys lost over ₹54,000 crore. The story everyone tells is simple and frightening: AI is going to eat Indian IT. Code that writes itself, testing that automates, legacy systems modernised by a model — fewer billable hours, smaller deals, shrinking margins.

It’s a clean narrative. The question that actually matters for your money is whether it’s true yet — or just feared.

The idea

The price fell on what investors expect; the business is so far moving on what clients actually do — and those two have separated. When price and fundamentals diverge this far, the gap itself is the opportunity or the trap. Naming which is the whole job.

What the fear says

The bear case is coherent. Generative AI can genuinely write code, handle large parts of software testing, and refactor old systems. Indian IT’s model has long rested on selling skilled human hours at scale. If a tool compresses those hours, the worry goes, billing shrinks, deal sizes fall, and the entire services pyramid narrows. Layer on US tariff noise and global capital being pulled into AI names instead of Indian equities, and you have a sector everyone wants to sell.

What the data says

Here’s the part the narrative skips. Look at the actual deals.

• Q4 results across the majors came in largely in line — revenue and margins held.
• Large-deal wins stayed healthy, often helped by GenAI, cost optimisation, and vendor consolidation.
• CLSA, examining Q4 deal renewals, found no evidence of AI-led pricing pressure. Pipelines intact. Demand commentary broadly stable.

So the thing the price is most afraid of — AI crushing deal economics — is not yet visible in the deal economics.

Why it matters

This is the cleanest example of a decision-intelligence call you’ll get this year. Two readings, opposite actions:

• Structural break: AI is early but inevitable; flat deals today are a lagging indicator; the de-rating is the market being right ahead of the numbers. Action: avoid, or trim.
• Sentiment overshoot: the business is fine, the fear ran ahead of the facts, and quality names now trade at undemanding multiples — TCS near a P/E of 17, Infosys 16, against years of double that. Action: this is where margin of safety shows up.

You don’t have to be certain. You have to size your bet to your conviction — and notice that the market briefly stopped distinguishing between AI will change this industry (true) and AI has already broken these companies (not yet shown).

How to think about it

1. Watch deal data, not headlines. Deal TCV, pipeline, pricing in the next two quarters is the real scoreboard. That’s where a structural break would appear first.
2. Separate the leaders from the laggards. The sell-off treated the sector as one trade. Analysts have already split — upgrading some, downgrading others.
3. Let valuation set your margin. At a P/E of 17, a quality compounder needs far less to go right than at 35. The price did you a favour even if the fear is partly real.
4. Define what would change your mind. Write down the deal-pricing number that would confirm the bear case. Then you’re reacting to evidence, not mood.

The takeaway

The market sold Indian IT for a future that hasn’t arrived in the numbers yet. Maybe it will. But “AI will reshape this industry” and “these businesses are already broken” are different claims — and this quarter the price treated them as the same. Spotting that difference is the edge.