The India global capability centre market has reached a size where the interesting question has changed. It is no longer whether to set one up. It is what you are setting up, because the thing being built in 2026 is not the thing that was built in 2016.
The nasscom-Zinnov India GCC Landscape report for FY26 puts the market at 2,117 centres operating across 3,728 units, with an estimated 2.36 million people and USD 98.4 billion in revenue. Centre count is up around 32% since FY2021, and 506 of the Forbes Global 2000 now have a presence.
Two of those numbers deserve more attention than the headline.
2,117 centres across 3,728 units
The gap between centres and units is the structural story. More than three and a half thousand units for just over two thousand centres means the average established company is not running one site — it is running a distributed footprint, usually across more than one city, often with different functions in each.
For a first-time entrant this reframes the location question. The decision is not "which city", it is "which city first, and what does that commit me to". Choosing a location optimised purely for the first fifty engineering hires can quietly rule out the finance or analytics function you intend to add in year three, because the talent pool, the salary benchmark and the attrition pattern are not the same.
2.36 million people, and what it does to retention
The headcount figure is labelled an FY26 estimate rather than a verified count, which is worth noting when it is quoted at you. The direction is not in doubt though, and the operational consequence is straightforward: in the established hubs you are hiring into a market where almost everyone you want already works for a GCC.
This changes what a value proposition has to do. Against 2,100 competitors in the same city, the differentiators that still work are the ones that are hard to copy quickly — the seniority of the work, whether decisions are made locally or escalated to headquarters, and whether the centre owns a product rather than a backlog. Compensation sets the floor. It does not win against a competitor who can offer ownership.
The three setup routes, and who each suits
| Route | What it is | Suits | Main risk |
|---|---|---|---|
| Direct | You incorporate, lease, hire and run it yourself | Companies with an existing India entity or a large, certain mandate | Slowest to first hire; you absorb every compliance and facilities decision |
| Build-Operate-Transfer | A partner builds and runs it, then transfers the entity to you | Companies that want ownership but not the first-eighteen-months learning curve | Transfer terms written vaguely at the start; agree them before signing, not at handover |
| Managed team | A partner employs the team; you direct the work | Testing a mandate, or a function too small to justify an entity | Less control over the employment relationship; harder to build a long-term identity |
The failure mode in BOT is specific and recurring: a transfer clause agreed in principle and left undefined. What transfers, at what valuation, on what notice, with which people and which contracts. A BOT without those four answers written down is a managed team with an aspiration attached.
What to decide before the first hire
- The mandate, in writing, with a sunset review. What this centre owns, what it does not, and when that is revisited. Centres without a written mandate default to whatever headquarters is willing to delegate, which is usually the least interesting work available.
- Reporting lines. Whether the India lead reports into a global function or into a country head determines, more than anything else, how senior the work becomes.
- The first ten roles. These set the ceiling. A centre started with ten junior engineers and one manager takes years to move upmarket; one started with three strong seniors moves faster and costs less in total.
- Statutory and payroll groundwork. Provident fund, professional tax, gratuity, shops and establishment registration, and state-level variation. Unglamorous, slow to fix retrospectively, and the single most common cause of a delayed first payroll run.
- An exit or restructure path. Rarely planned, occasionally needed, and far cheaper to design on day one.
The functions that have moved
Engineering and IT built the market, but the FY26 picture is broader: finance operations, data and analytics, cybersecurity, and increasingly AI and product roles. The implication for a first centre is that the talent you can realistically attract depends on what you are willing to move. A centre scoped as support will recruit against other support centres. A centre scoped to own a capability recruits against a much better field.
That is the real shift behind the numbers. The market has enough scale that candidates can choose, and what they choose on is the work. Coverage of the same report described India's centres as increasingly leading the AI mandate for their global parents rather than executing it — which is the clearest signal available about where the senior roles are going.
Key takeaways
- India hosted 2,117 GCCs across 3,728 units in FY26, with an estimated 2.36 million people — the footprint is distributed, so plan the second location while choosing the first.
- In established hubs you are recruiting from competitors, not from the open market; ownership of real work beats compensation.
- Pick the route — direct, BOT or managed team — on how certain the mandate is, not on headline cost.
- In a BOT, define what transfers, at what price, on what notice, and with which people, before signing.
- Write the mandate down. An undefined centre inherits the least interesting work available.