What the FY26 numbers show
The fiscal 2026 hiring data has one headline and one finding, and they are not the same thing. The headline is that AI hiring in India is growing. The finding is where it is growing, and at what level of seniority.
- GCCs now account for an estimated 30% to 35% of AI-related hiring in India — roughly a third of the market, sitting inside captive centres rather than supplier firms.
- They added nearly 200,000 net employees in fiscal 2026, almost twice the 110,000 added by IT services firms. That is the third straight fiscal year in which GCCs led net hiring.
- India had 2,117 GCCs employing about 2.36 million professionals in fiscal 2026, generating $98.4 billion in revenue. More than 1,200 of them have embedded AI and machine learning capabilities.
- More than 77% of GCC hiring is now mid-to-senior, according to TeamLease Digital, up from 60% in 2023, with demand strongest in AI, platform engineering, cloud, cybersecurity and data roles.
- Naukri JobSpeak recorded 33% year-on-year growth in AI/ML hiring in July 2026, the fastest-growing white-collar segment in the country.
Read those five lines together and the picture is not simply "AI jobs are up". It is that a specific kind of employer has become the default buyer of Indian AI talent, and that employer is buying experience rather than potential.
The centre of gravity moved, and most advice has not caught up
For twenty-five years the mental model of a technology career in India ran through services. You joined an IT services firm, you were staffed onto a client account, you moved up by managing larger accounts or larger teams. The advice ecosystem — placement cells, coaching, the entire apparatus of campus recruitment — was built around that path because that is where the volume was.
The volume has moved. Capability centres out-hired IT services roughly two to one in fiscal 2026, and it was not a one-year anomaly: it was the third consecutive year of the same pattern. A country with 2,117 GCCs and 2.36 million people working in them is not running a side channel any more. It is running the main channel, and more than 1,200 of those centres have AI and machine learning capabilities embedded in what they do.
That matters for a builder because the two employer types want visibly different things. A services firm is optimising for utilisation — billable people, deployed against contracts, with skills that transfer between accounts. A capability centre is optimising for ownership — people who can hold a system in their head for three years and be accountable for what it does in production. The second of those is much harder to staff, which is part of why the money has moved and why the AI talent gap now behaves like a supply problem rather than a demand one.
It is worth being precise about what that AI work actually consists of, because "AI role at a GCC" covers a very wide spread. At the centres doing it seriously, the job is rarely training foundation models. It is retrieval over the parent company's own document estate, evaluation harnesses that decide whether a model change is allowed to ship, guardrails and model governance that have to satisfy a regulator in a different jurisdiction, and inference cost control at volumes where a rounding error becomes a budget line. That is unglamorous next to research. It is also the work that compounds into a career, because very few people have done it at scale and the ones who have are visible.
It also explains why the frontier labs are turning up in person. When Anthropic opened a Bengaluru office, the pull was unlikely to be cheap engineering. India's GCC build-out has concentrated production AI teams in a way few other non-US markets have, and you do not open an office next to a labour arbitrage; you open one next to your users.
GCC, IT services or AI-native startup: what you are actually joining
The three employer types get lumped together as "tech jobs" and they are not the same job. The differences that matter to a builder are ownership, exposure to systems that real users touch, and what the interview loop actually screens for.
| Dimension | GCC (captive centre) | IT services firm | AI-native product startup |
|---|---|---|---|
| Typical pay band (India) | Top of the national range. The Rs 80 LPA-and-above senior figures come from GCCs and top product companies; 40-70% above services for equivalent experience. | The baseline the other two are measured against. Freshers cluster near the Rs 6 LPA floor of the 2026 range. | Cash bands overlap with GCCs — also 40-70% above services — with more of the total package typically in equity, which is not a number you can benchmark. |
| Seniority skew | Heavily mid-to-senior: more than 77% of fiscal 2026 hiring, up from 60% in 2023. | Still the widest entry door in the country by volume, though routine delivery work is also the most automatable part of the mix. | Small teams, senior-biased by necessity — there is rarely anyone spare to supervise a first job. |
| Scope of ownership | Increasingly global product and platform scope, not regional support. You own a system, not a ticket queue. | Scoped to the client contract. Ownership ends where the statement of work ends. | Total, and unbounded. You own the thing and everything adjacent to it that nobody else is holding. |
| Exposure to production AI | High and durable — evaluation harnesses, retrieval pipelines, model governance, cost control at real volume. | Variable. Depends entirely on the account you are staffed to; a bad draw can mean two years of maintenance. | Highest intensity, shortest history. You will ship a lot; you may not see a system age. |
| What they screen for | Evidence you have run something in production and can be trusted with the parent company's own roadmap. | Trainability, breadth, and the ability to be redeployed. Certifications carry more weight here than elsewhere. | Shipped artefacts. What you have built and can demonstrate, often before a formal interview loop begins. |
The honest summary: services is still the widest door for a first job, GCCs are where the durable AI platform work and the money now sit, and AI-native startups offer the most ownership per year served at the highest variance. There is a longer breakdown of how capability centres actually recruit in our guide to AI roles in Global Capability Centres across India and the UK.
Every article here is written by a Verified Builder. Want your name on the next one?
AI Tech Connect lists AI engineers, founders and researchers across India and the UK — and the people hiring browse it to find them. Adding your profile is free.
Become a Verified Builder →The mid-to-senior skew is the finding nobody leads with
Every hiring report this year has led with a growth number. TeamLease Digital's composition figure is the one that should have led instead: mid-to-senior talent rose from 60% of GCC hiring in 2023 to more than 77% in fiscal 2026. In three years, the share of intake available below that line fell by roughly half.
Put that next to the 33% year-on-year growth in AI/ML postings that Naukri JobSpeak recorded in July 2026 and you get two facts that are usually presented as one. Postings grew fastest of any white-collar segment. And the employers doing most of that hiring are buying experience. Both are true, and only the first one ever makes the headline.
If you are a fresher or a career-changer, do not read 33% posting growth as an open door. The segment growing fastest is the one skewing hardest towards mid-to-senior, and applying into it with no production evidence is how people end up concluding the market is closed when it is simply stratified. The realistic first move is a role that puts you next to a production AI system — including at a services firm — and a public record of what you shipped there.
"The pattern I keep seeing in Chennai and Bengaluru is that capability centres are not really running graduate funnels for AI roles — they are sourcing people who already have two or three years of production scar tissue, often out of services firms. That is a genuine path, but it is a two-step one, and nobody tells you that when you are starting out."
— Rishi, Verified Builder · Chennai, IndiaThere is a structural reason for the skew. A capability centre exists to take ownership of systems the parent company depends on. Ownership is not something you can hand to a first-year engineer, and unlike a services firm, a GCC has no bench economics that make training large cohorts pay for itself. So it buys people who are already finished. That is a rational choice for the employer and an uncomfortable one for anyone standing at the entrance.
What this looks like from the UK
The instinct from a UK vantage point is to convert the Indian figures into pounds and conclude that nothing here is relevant. That comparison is real, and it is also the least interesting one available.
| Measure | India | UK |
|---|---|---|
| Open-role volume | No single directly comparable count. Naukri JobSpeak recorded 33% year-on-year growth in AI/ML hiring in July 2026, the fastest-growing white-collar segment. | Roughly 1,465 open AI engineer roles listed in London as of August 2026 — a job-board count from Glassdoor, not an official statistic. London leads the non-US hubs on volume. |
| Median pay | No published national median for the role. The 2026 range runs roughly Rs 6 LPA for freshers to Rs 80 LPA or more for seniors. | Around £56,614 for an AI engineer, within a range of roughly £32,461 to £102,496. |
| Senior band | Rs 80 LPA and above at GCCs and top product companies — roughly £70,000 at current rates, and a small number of roles rather than a broad band. | £90,000 to £150,000 base for seniors, before equity or bonus. |
| Currency-normalised read | Pound-for-pound the UK wins at every level, and nothing in the FY26 data suggests that gap closing soon. But London leads its peer hubs on volume at materially lower pay than the US, so British builders are already living on the wrong side of one of these comparisons. The variable worth tracking in both markets is scope per unit of pay, not pay alone. | |
Two things follow for a UK-based builder. The first is that a growing share of the AI platform work inside large multinational employers is now designed, built and operated from Indian capability centres — which means the counterpart team on your next project may well be a GCC team, and the interesting question is who owns the architecture rather than who is cheaper. The second is that scope is the currency that transfers. A GCC engineer owning a global retrieval platform accumulates a stronger record than a London engineer maintaining one regional service on twice the salary, and on current trends the market looks set to price the record rather than the postcode. Before either side negotiates, it is worth benchmarking the two markets properly rather than by anecdote.
The logic also cuts both ways as a market. The UK hosts capability centres of its own, and the same test applies to them: the roles worth chasing are the ones carrying a global mandate rather than regional support, and they are screened on the same evidence. A London-based builder reading these Indian figures should take a method from them rather than a destination — find the employer type that owns systems instead of tickets, then judge the offer on what you will actually be accountable for.
None of that makes the comparison like-for-like. It makes it a different comparison from the one people usually run.
What a builder should do about it
The practical translation of a market that is 30% to 35% GCC-driven and more than 77% mid-to-senior is fairly specific.
If your goal is depth and durability — owning a system long enough to see it break, age and get rebuilt — target GCCs. That is where the AI platform work with a multi-year horizon sits, and it is where the top of the Indian pay range lives. Expect to be screened on production evidence rather than potential.
If your goal is breadth or a first job, services remains the widest entry door in the country. The trick is to treat it as a two-year evidence-gathering exercise rather than a career track: get onto an account with real AI delivery, and document what you personally owned.
If your goal is ownership per year served, an AI-native product startup gives you more of it than either alternative, at higher variance and with a package you cannot benchmark. The due diligence burden is on you, and it is worth doing properly before you sign.
If you are reading this from the UK, the same composition finding cuts differently. The employer archetypes are not the Indian three, but the underlying shift is: the AI platform work inside large multinationals is increasingly specified and operated from capability centres, so the roles worth chasing in London, Manchester or Edinburgh are the ones that own architecture rather than regional delivery. Screen a UK offer on the same axis an Indian builder screens a GCC on — how much of the system will you actually own, and for how long — rather than on the salary line, which already favours you.
When a market skews mid-to-senior, recruiters shift from sifting applications to sourcing candidates directly — because the people they want are not applying. That inverts what makes a difference. A polished CV only helps once you are already in a pile; a public, findable record of shipped work is what gets you into consideration at all. Make sure there is a page with your name on it that shows what you have built, what it runs on, and what it does in production.
That last point is the one that has changed most in the past three fiscal years, and it is easy to miss because it is a change in method rather than in numbers. When capability centres were hiring in bulk at entry level, applications worked. Now that more than three-quarters of their intake is experienced, the pipeline runs the other way. Being findable is no longer a nice-to-have for the ambitious; it is increasingly how these roles get filled.
The source reporting on GCC and cloud hiring across APAC is at TechRepublic, and Naukri publishes its monthly JobSpeak index on the Naukri blog. Both are worth tracking monthly rather than annually — the composition of hiring is moving faster than the volume.