MARKET COMMENTARY · IT SERVICES · GLOBAL CAPABILITY CENTRES
MARKET COMMENTARY · July 2026 · BY PIVOTAL RESEARCH
India's software exports are still growing, but the listed IT services majors are barely moving. The missing growth is increasingly being absorbed by Global Capability Centres, AI, and a new client-owned operating model.
For most of the last thirty years, the story of Indian technology was a single story. Global companies had work to do, Indian firms had the talent to do it, and TCS, Infosys, Wipro and HCL built enormous businesses standing in the middle. That middle is now the problem.
Start with a puzzle. India's software and services exports are still growing at a healthy clip. Yet the listed IT majors, the traditional heavyweights of that export engine, are barely moving. Infosys closed FY26 with revenue growth of 3.1% in constant currency, and guided to roughly 3% to 3.5% for the year. A decade ago, growth in the high single digits or low double digits was routine. Emkay Global's analysis of the top three firms notes that in the pre-pandemic years of FY17 to FY20, constant-currency revenue grew about 10% a year. The BSE IT index fell around 14% over the past year even as the broader market held up.
Something is absorbing the growth that used to flow to the big vendors. A large part of the answer is the Global Capability Centre.
The scale of the shift
A GCC, sometimes still called a captive or a global in-house centre, is a technology and operations hub that a multinational owns and runs itself, rather than outsourcing the work to a vendor. India now hosts the overwhelming majority of them.
The numbers are large and moving fast. NASSCOM and Zinnov put Indian GCC revenue at USD 64.6 billion in FY24, with more than 1,700 centres employing around 1.9 million people. By FY26 industry trackers estimate the market at close to USD 100 billion, across more than 2,100 centres. The joint NASSCOM-Zinnov projection for 2030 is USD 99 to 105 billion in revenue, 2,100 to 2,200 companies running well over 4,000 individual units, and a workforce of 2.5 to 2.8 million. India already accounts for a little over half of all GCCs worldwide.
To put that in context, NASSCOM expects the entire Indian tech sector to cross USD 300 billion in FY26. GCCs are now close to a third of that and climbing faster than the services firms they increasingly compete with.
The people who build these centres for a living think the shift has barely begun. Lalit Ahuja, founder and chief executive of ANSR, which has set up well over a hundred captives in India, argues the sector is only "at the end of the beginning," with the next decade set to bring exponential growth. His explanation for why the work moved in-house is worth sitting with: companies came to see technology not as a cost to be outsourced but as a core differentiator to be owned, and once that realisation set in, they invested heavily in building the capability themselves rather than renting it from a vendor.
The case that GCCs are cannibalising IT services
The clearest evidence is geographic. The United States, the single largest market for Indian IT, saw revenue growth for the major firms slow to roughly 1% to 2% in FY24 and FY25. That is not a normal cyclical dip. It points to a structural change in how American companies get their technology work done, and the change is that they are increasingly doing it themselves, in India, through their own centres.
The talent flow tells the same story. The people who staff these captives are, very often, the same engineers who once worked for the vendors. Media estimates suggest that across India's five largest IT services firms, more than 80,000 roles disappeared in the eighteen months to mid-2025. TCS reduced its workforce by around 12,000, close to 2% of its global headcount, in a single restructuring in 2025, its largest such reduction in memory. By one Economic Times count, 2024 and 2025 together saw close to 99,000 technology job losses across more than 330 Indian companies. Not all of this is GCC-driven, and not all of it is distress. A meaningful share is simply the client hiring the talent directly and no longer needing the intermediary.
Most damaging of all is which work is moving. In the early years, captives handled the routine and the repetitive while the interesting work stayed with vendors or moved onshore. That has inverted. Digital strategy, research and development, AI, and product ownership are increasingly being built inside the GCC. NASSCOM's own data shows nearly half of Indian GCCs now manage end-to-end product lifecycles, and engineering-focused centres are growing faster than the GCC average. This is the high-margin work that the services firms most wanted to keep.
The offshore cost advantage Indian vendors once sold was "a differentiation Indian service providers had" that their own clients can now capture directly through a captive, disrupting the classic offshore development centre model.
Pari Natarajan, Chief Executive, Zinnov
Emkay Global captured the irony of the moment in four words. The firms that spent decades disrupting Western in-house IT departments through offshoring are now on the receiving end of the same logic. In its words, "the disruptor will get disrupted."
The fight for people, and how AI rewrote it
If you want the mechanism of the cannibalisation in one word, it is talent. A GCC does not win by out-marketing a vendor. It wins by hiring the vendor's best engineers and handing them better work.
The pay gap is the blunt instrument. GCCs hold a 15% to 22% salary premium over the IT service providers, according to Zinnov, and for a decade they used it to poach aggressively. But money was never the whole story. The sharper draw is the work itself. As architecture decisions, product roadmaps, and even pricing and profit-and-loss responsibility moved into the India centre, the interesting problems moved with them. Attrition at the vendors rose less because people were unhappy and more because better work now existed a short commute away.
That talent war, in its old form, is cooling. Zinnov describes the shift as a move "from salary race to skills chase": the bidding wars of the Great Resignation giving way to a measured, skills-first market. Nearly half of GCCs now say they hire on proven capability rather than on credentials. Overall attrition has drifted down from 19.4% in 2021 to around 17.4% in 2024. Hiring has turned lateral and deliberate rather than bulk.
But cooling is not the same as solved, and this is where the honest picture gets uncomfortable. The shortage did not close. It moved upmarket. India now runs an estimated 90% shortfall in generative-AI-ready talent, a 55% to 60% deficit in cloud-native skills, and a 25% to 30% gap in mid-to-senior cybersecurity, on the numbers circulating from NASSCOM, Zinnov and EY. Demand for generative-AI specialists has jumped roughly 300%, the salary premium for them reaches 60%, and some searches now run 90 days. The most acute gap of all is leadership: the professional with eight to fifteen years of experience who can combine technical depth with the judgement to run a strategic centre. As NASSCOM's own talent playbook puts it, the contest is now about "quality, relevance, and speed", not headcount.
Then AI arrived at the hiring process itself, and rewrote its rules from the bottom.
For thirty years the Indian model rested on one mechanism. Hire freshers in tens of thousands, park them on the bench, train them on testing, documentation, ticket summaries and basic code, and grow them into billable engineers. That is precisely the work generative AI now does. So the bottom rung of the ladder is being pulled up behind the people already standing on it. TCS extended around 25,000 campus offers this cycle, down from the 40,000-plus it once made. Direct campus hiring across the sector is running 30% to 35% below its historical average, and EY estimates the IT firms have cut entry-level roles by 20% to 25%. Entry-level openings fell 44% year on year by the middle of 2026. In one global survey, 37% of managers said they would rather use AI than hire a recent graduate.
The uncomfortable part, for anyone hoping the GCC boom simply absorbs those displaced freshers, is that it does not. GCC hiring skews hard to the four-to-ten-year specialist, and roughly 64% of new GCC roles now demand AI, data, or automation skills that a fresh graduate does not yet have. The centres are growing: GCC openings rose even in months when the broader market fell. But they are growing at the top of the skill curve, not the bottom. NASSCOM expects more than 1.5 million Indian IT roles to be significantly transformed by AI by 2027.
The net effect is a barbell. The middle of the experience curve, the immediately useful engineer with four to ten years behind them, has never been more sought after or better paid. Both ends are being squeezed: the fresher who used to learn on the bench, and increasingly the expensive generalist manager whose span AI is starting to compress. The talent that both the GCC and the algorithm now reward is narrow, specific, and in short supply. Everyone else is discovering that the pipeline they were standing in has quietly been taken away.
Why the reality is more complicated than a straight substitution
It would be easy to stop there and declare the IT services model dead. The honest picture is more layered, for three reasons.
First, the market is growing, not fixed. This is not a pure zero-sum transfer where every rupee a GCC earns is a rupee taken from a vendor. Global technology budgets are expanding, India's total tech exports are still rising, and both GCCs and services firms can grow at once. What is really being destroyed is a particular model, the labour-arbitrage and staff-augmentation business of putting bodies on seats, rather than IT services as a category.
Second, AI is a simultaneous disruptor, and arguably a larger one. The same period that saw GCCs surge also saw the arrival of generative and agentic AI, which compresses exactly the kind of routine engineering and support work that filled the bottom of the services pyramid. Separating the GCC effect from the AI effect is genuinely hard, and much of the commentary that blames GCCs is partly describing AI. TCS and Infosys both now frame their restructuring around a pivot to AI-first delivery, not around GCC competition.
Third, and most interesting, the vendors are turning the threat into a product.
The counter-move: selling shovels in the gold rush
If clients want to build their own centres, the IT firms have decided to help them do it, for a fee. Almost every major vendor now has a GCC-as-a-Service offering that builds, staffs, runs and eventually hands over a client-owned centre.
Infosys has mandated dedicated GCC offerings with embedded AI across its delivery network and is setting aside campus capacity to incubate client-owned hubs. HCLTech and Tech Mahindra have opened units focused on setting up and running captives. LTIMindtree launched a build-operate-transform-transfer service priced per seat. Wipro and Cognizant have similar plays, and mid-tier firms such as Mphasis and Hexaware are investing in the advisory firms that help corporations stand up centres.
The logic is uncomfortable but rational. As one executive with knowledge of Infosys's strategy described it, the firm is "cannibalising its own IT revenue to grab more market share amongst GCCs." Better to earn the enablement fee than to lose the client entirely. LTIMindtree's chief executive, Venu Lambu, frames the same shift more positively, describing GCCs as "strategic centers for industry-specific transformation and efficiency" that the firm can help build. Some industry estimates suggest that by 2028 a large share of IT services revenue could come from GCC-related partnerships rather than from competing against them.
The twist: the captives are not safe either
There is a final irony worth holding onto. The GCC model is not a permanent fortress. The same AI wave that is hollowing out the services pyramid is now aimed at the captives.
Zinnov and Indiaspora estimated in early 2026 that around 55% of the work currently done inside Indian GCCs sits in commoditised, procedural tasks that are vulnerable to automation. Natarajan's warning about the captives was blunt: the portfolio of work that justifies a centre's existence is, in his phrase, "now the most exposed layer." A GCC that exists mainly to run repeatable processes at lower cost is exposed to exactly the same disruption it was built to deliver.
So, are GCCs eating IT services?
Yes, but that is the wrong frame if it is the only one.
The precise answer is that GCCs are eating a specific, once-lucrative layer of the IT services business: the arbitrage model, the offshore development centre, the practice of renting out trained engineers to clients who no longer wish to rent. On that count the cannibalisation is real, it is structural rather than cyclical, and the divergence between booming software exports and stagnant vendor revenue is the proof.
But the same layer is also being eaten from below by AI, and the surviving vendors are quietly repositioning as the builders and operators of the very centres that threaten them. The commodity middle, the part of the industry that simply moved labour from an expensive geography to a cheaper one, is being squeezed from both directions at once.
The winners of the next decade, whether they carry a vendor's logo or a multinational's, will be the ones who own genuine capability: deep domain expertise, real product ownership, and the ability to deliver outcomes rather than hours. The losers will be whoever is still, in 2030, mainly selling the arbitrage that both the GCC and the algorithm have made abundant.
Sources: NASSCOM-Zinnov India GCC Landscape reports; NASSCOM sector estimates; company results (TCS FY26, Infosys FY26); Emkay Global and CRISIL sector analysis; Zinnov (salary, attrition & hiring trends) and ANSR commentary; NASSCOM talent playbook; EY, Deloitte and foundit hiring data; Business Standard, Economic Times, Mint, The Ken, Rest of World and HRKatha reporting. Figures reflect the most recent public data available and vary by source and methodology.
-- PR --