TCS Cut 23,000. Infosys Hired 20,000. Who's Right?
TCS cut 23,000 jobs while Infosys hired 20,000 freshers. The real Indian IT AI workforce split is subtler, and FY27 will show whose bet pays off.
Note: Statistics and figures reflect data available as of late September 2026, drawn from company earnings calls, filings and contemporaneous reporting. Verify for latest figures.
Two numbers from FY26 have been doing the rounds all year.
TCS ended the year with 584,519 employees, down from 607,979 a year earlier. That is a net drop of about 23,460 people, per Business Today's report on the Q4 results. In the same year, Infosys "onboarded more than 20,000 freshers," CFO Jayesh Sanghrajka said on the Q4 FY26 earnings call, and said it expects to hire a similar number in FY27.
Put those side by side and you get an easy story: one company read the AI shift and cut, the other kept hiring, so one must be wrong.
I think that story is mostly wrong, and the way it's wrong is more useful than the headline. Both companies are making a different, testable bet about where the value of AI-assisted delivery will sit. FY27 results will show which one works, so this post writes down now what to look for.
The headline comparison hides the real split
Start with the detail that ruins the tidy version. TCS did not stop hiring freshers. Its CHRO Sudeep Kunnumal said, per the same Business Today report: "Last year, we onboarded 44,000 trainees. We have already made 25,000 campus offers in India."
That is more than twice Infosys's fresher intake. It is bigger relative to size too. 44,000 trainees is about 7.5% of TCS's year-end headcount. Infosys's 20,000 is about 6.1% of its 328,594.
Infosys's net number is also less generous than "hired 20,000" suggests. Its 20-F filings show headcount going from 323,578 in March 2025 to 328,594 in March 2026. That is a net gain of about 5,000 people, or roughly 1.5%. In the March quarter alone, headcount fell by 8,440.
Then the June quarter flipped the headlines. TCS added 9,279 employees net, its biggest quarterly addition in about four years. On the Q1 FY27 call, Infosys said headcount "reduced by 500 employees after adding over 2,000 employees from acquisitions," even as it had recruited "over 4,000" graduates in the quarter.
So the real difference is not "cut vs hire". Both companies run a big campus funnel. The difference is what they did to the layers above it, and what they say AI will do to those layers.
Bet A: TCS thins the middle, not the base
The restructuring that started all this was announced in July 2025. TCS said it would cut about 2% of its workforce, around 12,000 roles, mostly in middle and senior grades. It called this "future-ready transformation." The actual FY26 drop ended up at nearly twice that. The company's position is that the extra drop was not all restructuring, and that the layoff cycle is over.
My reading of the bet: AI makes the coordination layer thinner. The classic pyramid had many layers of people who planned, reviewed, reported and escalated work. If agents do a growing share of the routine build and test work, and tooling handles more of the status tracking, you need fewer of those layers. You keep hiring at the base because young engineers are cheap and can be trained on AI-native workflows from their first day. What you stop funding is the thick middle.
CEO K. Krithivasan has been careful to frame it as productivity, not shrinkage. After the Q1 FY27 results he said TCS does not expect AI to reduce its overall workforce. He said the efficiency gains will go into running more projects, not into cutting jobs. The same report puts TCS's AI portfolio at an annualised revenue run rate of $2.6 billion.
What has to be true for Bet A to work:
- Juniors plus AI tools really can do work that used to need a mid-level engineer checking it. If not, quality problems show up as rework and client escalations.
- The seniors who stayed can supervise a wider span of people. Thinner middles only work if the remaining managers can handle more direct reports and more agent-driven output.
- Clients accept the new shape. A client who is used to a named delivery manager and a named module lead may not like the flatter version, even if it delivers.
The risk is subtle. Middle layers are where you learn to become senior. Cut too deep and in five years you have plenty of juniors, a thin layer of very senior people, and nobody in between who is ready to step up.
Bet B: Infosys keeps the funnel and retrains around it
Infosys's bet looks more conservative, but it has a specific shape. Salil Parekh said on the Q1 FY27 call that new graduates now arrive with a "native understanding of AI". Infosys then trains them on its own AI platforms (Topaz and Fabric). He also laid out a plan for "6,000 frontier engineers over the next few years." Infosys also reported AI services at 8.2% of revenue, growing at double digits quarter on quarter.
The bet: AI-trained freshers become the new base of the pyramid, and the base stays wide. Instead of taking out a layer, Infosys keeps its structure and tries to change what each layer does. Utilisation excluding trainees rose to 84.9% in Q1 FY27, per the same call. That suggests the bench is already tight, so the new hires are there to meet demand, not to sit idle.
What has to be true for Bet B to work:
- Demand recovers enough to put 20,000 new people on billable work. Infosys trimmed its FY27 revenue guidance to 1.5-3% in constant currency on that call. That is not a lot of room.
- The retraining actually changes output per person. A wide base that does the same work at the same speed just moves the old cost structure into a year of AI-driven price cuts.
- The middle layer can absorb change without a layoff. Infosys has avoided TCS-style headline cuts. If revenue stays flat, that kind of restraint gets expensive.
What Infosys keeps that TCS risks is the pipeline to senior roles. What it risks is carrying a cost base built for a growth rate it has not had for two years.
Wipro: what hedging costs
Wipro is the useful third case because it did not really choose. It hired about 7,500 freshers in FY26, over 3,000 of them in the March quarter. Then it declined to set any FY27 target. "We don't have any target for fresher hiring for the next fiscal. It's completely on demand, very volatile environment," CHRO Saurabh Govil said, per Business Today.
The cost of that caution lands on people who already said yes. Mint reported in September that about 400 trainees in Wipro's WILP work-and-study programme now face permanent conversion around March 2027. They had originally expected it in December 2024. The same report says more than 200 candidates from its Elite hiring process were still waiting to be onboarded. Business Standard carried the same story.
My point isn't that Wipro was careless. It's that hedging is not free. If you keep offers open while deciding whether you believe in the pyramid, the uncertainty lands on 22-year-olds who turned down other jobs. A clear bet in either direction at least tells people where they stand.
The industry twist: total headcount still grew
Now zoom out, because the company numbers leave out where most of the hiring actually happened.
Nasscom's Strategic Review 2026, released in February, estimated that India's tech workforce would reach almost 6 million (about 59-60 lakh) in FY26. That is a net addition of about 135,000 people, or 2.3%. These were full-year estimates made before the year closed. Treat them as Nasscom's projection, not a final count.
So the sector added about 1.35 lakh jobs in a year when the biggest employer shed more than 23,000 and the second biggest added about 5,000. Much of the difference is global capability centres (GCCs), the in-house tech hubs that multinationals run in India. The same Nasscom coverage puts India at more than 1,700 GCCs employing around 1.9 million people. I wrote about why they're growing in the GCC piece.
That changes the question. If a bank moves work from a TCS contract to its own Bengaluru centre, the engineer's job didn't disappear. It moved to a different payroll. Some of what looks like "AI cutting services jobs" is really "clients insourcing the work AI makes easier to manage in-house."
The counterargument: maybe neither is an AI strategy
The strongest objection to this whole post is that I'm giving both companies too much credit.
TCS's revenue went slightly backwards in FY26. Infosys guided to low single digits. When demand is weak, you cut the most expensive layer and slow hiring. When it picks up, you hire again. TCS's Q1 FY27 rebound came alongside a better order book. That is exactly what an ordinary demand cycle looks like, and "AI transformation" is a much better label for it than "we had a soft year."
I've gone through this argument for FY26 revenue per employee in a separate post, so I won't repeat it all. The short version: much of the FY26 ratio improvement came from clearing benches and pyramid pressure that predate agents. The same caveat applies here. Some of each "bet" is simply how each company responds to a downturn, described in AI terms.
Where I come down: the demand-cycle explanation covers most of the FY26 timing. It doesn't explain the shape. TCS cutting specifically in middle and senior grades, and Infosys putting a number on frontier engineers and AI revenue share, are choices. A pure cyclical response would have cut evenly or frozen hiring altogether. My read is that both are cyclical responses with a real strategic lean, and the lean is what FY27 will test.
The scorecard for April 2027
Here's what I'll check when FY27 annual results come out, written down now so I can't quietly move the goalposts.
| Metric | What favours TCS's lean bet | What favours Infosys's wide bet |
|---|---|---|
| Revenue per employee | Rises again with headcount flat or up | Holds steady while headcount grows |
| Utilisation (excl. trainees) | Stays high without a big bench | Stays near 85% despite 20,000 new hires |
| Fresher-to-lateral mix | Freshers stay high, laterals fall (middle stays thin) | Both steady; the pyramid shape holds |
| Disclosed AI revenue | TCS's run rate grows faster than headcount | Infosys's 8.2% share keeps climbing and stays reported |
| Attrition | Doesn't spike as seniors leave for GCCs | Stays near 13% despite slow growth |
Two tests matter most. First, does TCS's thinner middle show up as client escalations or margin leakage? You'll hear about that on calls as "delivery investments." Second, does Infosys keep reporting AI revenue as a separate line? A metric that quietly disappears usually means it stopped looking good.
If I had to bet, I'd put slightly more on TCS's version being where the industry ends up. I'm not confident, though, and I'd put the Infosys side at not much below even. The lean model fits what clients are asking for in pricing talks. But it has a talent-pipeline cost that won't show up for three or four years, long after FY27 results.
What it means for you
If you're a fresher: the good news is that both of the biggest firms are still hiring at the base. The bad news is that the jobs you'd have grown into, module lead and delivery manager, are the ones under pressure. Plan for a career where you move sideways into judgement-heavy work sooner. That means reviewing AI output, owning a client domain and debugging systems you didn't build. The freshers post covers what that looks like in practice.
If you're mid-level: you are exactly the layer this whole debate is about. At a TCS-style firm, your safety comes from being the person who makes a flatter team work: wider span, more agent-driven output, fewer handoffs. At an Infosys-style firm you have more time, but not unlimited time. Either way, GCCs are the obvious lateral move, and the Nasscom numbers say they are hiring.
If you're a manager: the question to ask yourself is whether you could run your current scope with one fewer layer under you. If the honest answer is yes, someone above you has probably asked it too. If the answer is no, be ready to explain why in terms of risk and client outcomes, not headcount.
If you want to look at your own exposure instead of the industry's, the AI job risk calculator on this site asks the same question at the level of an individual role. It separates the parts of your job that are genuinely automatable from the parts that just sound like they are.
Related Reading
Enjoying this article?
Get posts like this in your inbox. No spam, unsubscribe anytime.
