The One Number Indian IT Never Moved Just Moved
Revenue per employee rose 3-4% across India's top IT firms in FY26, after two flat decades. What actually caused it, and what FY27 has to prove.
Note: Statistics and figures reflect data available as of September 2026. FY26 figures are drawn from company results and contemporaneous reporting; verify against the latest filings before relying on them.
For twenty years, Indian IT had one number that never really moved: revenue per employee.
It drifted. It got a point or two better in a good year and gave it back in a bad one. But structurally it was flat, and that flatness was the business model. Revenue grew because headcount grew. If you wanted 10% more revenue, you hired roughly 10% more people, trained them in a bench-funded pipeline, and slotted them into a pyramid. Everything about the industry — campus recruiting, utilisation targets, the pyramid itself — was built on the assumption that output scales with bodies.
In FY26, that number moved. Per Angel One's analysis of the FY26 results, TCS, Infosys and HCLTech each posted a 3-4% rise in revenue per employee — TCS at about 3.4% and HCLTech highest at roughly 4.1%. Wipro went the other way and declined.
Three or four percent doesn't sound like a revolution. Hold that thought, because where it came from is more interesting than how big it is.
First, the part everyone gets wrong
The headline you've probably seen is some version of "Indian IT is earning more from fewer people." It's the tidy story, and it's not quite what the filings say.
FY26 revenue growth was wildly uneven across the tier-1s. Per The Register's April 2026 round-up of the FY26 results, HCLTech grew about 11.2% while increasing headcount roughly 2%. Wipro grew around 4%. Infosys came in around 3%. And TCS — the largest of them — posted a revenue decline of about 0.5%, which CEO K. Krithivasan acknowledged on the call as degrowth.
So no, it isn't "everyone earned more with fewer people." One firm grew double digits while hiring. One shrank.
The thing they share is the ratio, not the direction. And that's what makes it worth writing about, because a rising revenue-per-employee number at a company whose revenue fell can only have come from one place: the denominator.
TCS improved revenue per employee by 3.4% in a year its revenue went slightly backwards, while its workforce shrank by about 23,000 people, nearly 4%, per its own results. That is not a productivity story dressed up in a headcount story. It is a headcount story, full stop.
HCLTech is the genuinely interesting one — the highest RPE gain in the group, achieved while adding people. If AI-driven delivery productivity is real anywhere in these numbers, it's there.
What's actually driving it — three causes, routinely conflated
Almost every article on this conflates three separate mechanisms. They have very different implications, so it's worth pulling them apart.
1. Genuine AI-driven delivery productivity. Real, and now showing up in disclosure. Infosys CEO Salil Parekh said AI work reached 8.2% of total revenue in Q1 FY27, growing at double digits quarter on quarter, per the company's Q1 release. A year ago that work would have been buried inside "digital." The fact that it's now a named line item is itself the signal — you don't break something out until it's big enough to be asked about.
2. The post-2022 over-hiring correction finally clearing. The industry hired aggressively into the pandemic demand spike and spent three years working it off. Benches are much thinner than they were in 2022, and utilisation is correspondingly higher. Higher utilisation raises revenue per employee arithmetically, with no change in how productive anyone actually is. This looks identical to productivity in the ratio and is completely different in nature.
3. Pricing and pyramid pressure that predates AI. Clients have been squeezing rates and pushing back on pyramid-heavy staffing since well before agentic coding existed. Leaner pyramids raise average revenue per head because you removed the cheapest heads.
My honest split, and I'll own it as a judgment rather than a finding: for FY26 specifically, cause 2 is doing most of the work, cause 3 is doing a solid chunk, and cause 1 is real but smaller than the commentary suggests. The bench clearing is a one-time effect that mechanically flatters the ratio, and it happened to land in the same year everyone was already primed to attribute everything to AI.
That's not scepticism about AI's eventual impact. It's scepticism about this year's number as evidence for it.
Why I still think the decoupling is real
Two reasons, both forward-looking rather than in the FY26 data.
The first is what executives are guiding to rather than reporting. HCLTech's C. Vijayakumar has been unusually direct about "AI deflation": he flagged a 3-5% hit to FY27 from AI-driven efficiency as clients capture the savings, per The Register. Think about what it takes to say that out loud. You don't warn the market about deflation you expect to be temporary. That's guidance about a structural change in what work is worth.
The second is that the bench-clearing explanation has an expiry date. You can only clear the bench once. Utilisation that is already high doesn't have much room left. If FY27 shows another 3-4% RPE gain, it cannot come from the same place, because the same place is empty.
Which sets up a clean test, and I want to state it as a falsifiable prediction rather than a vibe:
If FY27 headcount stabilises while revenue-per-employee keeps climbing, the AI productivity story holds. There's nowhere else for the gain to come from.
If revenue flattens and headcount keeps falling, this was a demand downcycle wearing an AI costume — a cost-cutting exercise that got a flattering narrative attached because the timing was convenient.
I'd bet on the first, but not heavily — maybe 60/40. Anyone more confident than that in either direction is selling something.
Three things worth watching in the FY27 numbers, because they separate the two stories cleanly. Utilisation: if it stays flat while RPE climbs again, the gain isn't coming from the bench. Fresher hiring: a genuine productivity shift lets you resume campus intake, because you're not trying to hold the ratio down artificially; a cost-cutting exercise keeps the tap shut. And the AI revenue line itself — Infosys has now put a number on it in public, which means it has to keep appearing quarter after quarter. A disclosure that quietly gets folded back into "digital" would tell you a great deal.
What it means if you work in this industry
If you're a fresher or near the base of the pyramid: the honest read is that the part of the structure that shrank is the part you were meant to enter through. That's not a temporary hiring freeze, it's a change in shape. The traditional path — join at the base, learn on a large team, move up as the pyramid grows under you — assumed a pyramid that keeps widening. It isn't widening.
The practical implication isn't "learn AI." Everyone says that and it's nearly useless as advice. It's that the work which used to be your training ground — the routine, well-specified, high-volume tasks — is exactly the work agents do acceptably now. So the skills that used to come after two years of that grind are the ones you now need on arrival: reading a codebase you didn't write, judging whether output is correct, and working out what the client actually meant. Those are reviewing skills, and they're now entry-level requirements.
If you're mid-level: you are the beneficiary and the target simultaneously. Higher revenue per employee means the firm needs fewer of you, but values each one more. High utilisation means the bench that used to absorb a slow quarter isn't there. That's a real increase in individual exposure, and it's worth planning around rather than being surprised by.
If you're a manager: your estimation models are wrong and your headcount plan is a worse proxy for capacity than it has ever been. I've written before about what agentic AI does to services margins; the FY26 numbers are the first year of that showing up in results rather than commentary. If you're still sizing engagements by person-months, you're pricing a unit that your client no longer believes in.
What it means for the firms
The pivot playbook hasn't changed much from what it looked like a year ago — outcome-based pricing, AI governance and audit services, orchestration as a managed offering, proprietary accelerators instead of pure labour. What FY26 adds is a year of evidence about which parts are working.
HCLTech's combination — double-digit revenue growth, headcount up, best-in-group RPE gain — is the only one in the set that looks like genuine operating leverage rather than cost discipline. Whatever they're doing is the thing to study, and it's notably not "shrink the workforce."
The uncomfortable version, which I think is true: the firms that improved the ratio by cutting have bought themselves a year. The ratio improves once when you clear the bench. It improves permanently only if the work itself changed.
FY27 tells us which one happened.
If you want to think about your own exposure rather than the industry's, the AI job risk calculator on this site walks through the same question at an individual level — which parts of a role are actually automatable versus which just sound like they are.
Related Reading
Enjoying this article?
Get posts like this in your inbox. No spam, unsubscribe anytime.
