Set Up a GCC in India Without Waiting for an Entity

A global capability centre is a company's own operation in India: its people, its entity or a structure heading toward one, working exclusively on its roadmap. The rebrand from captive centre to GCC is not empty. The old captive did cost arbitrage on support functions; the 2026 GCC ships core product, owns platforms, and increasingly houses the functions a company would once have guarded at headquarters: data science, security operations, finance transformation, product management.

India hosts over 1,700 of these centres employing close to two million people, and the interesting shift is who builds them now. It used to be Fortune 500s with consultants and five year plans. It is now 200 person startups opening 15 person pods, because the machinery for starting small finally exists.

📇 What a GCC is not

It is not outsourcing. An outsourcing vendor's staff work for the vendor, on the vendor's processes, for many clients; a GCC's people work for you alone, in your tools, carrying your culture. It is also not defined by a building or a legal structure. A 12 person pod employed through an employer of record, working exclusively on your product with your rituals, is functionally a GCC in miniature. The definition that survives contact with 2026: exclusive dedication plus your management, whatever the paper says.

🤔 Why the definitions fight matters

Because vendors price the label, not the substance. Call it a GCC and consultants quote transformation programmes; call it a pod and the same team costs a flat monthly fee per head. Through this article we will hold one test: who employs the people, who directs the work, and who keeps the capability if the vendor disappears. Those three questions cut through every brochure in this market.

For the employment layer that sits underneath every structure in this piece, our EOR services in India page is the reference: it is the machinery this article keeps pointing back to.

Every GCC gets built one of three ways. The brochures make them sound like a maturity ladder. They are actually three different bets about speed, control and reversibility.

🚧 The BOT fine print

Build-operate-transfer sounds like renting with an option to buy, and sometimes is. The clauses that decide whether it was: the transfer fee, often priced per transferred head at the moment you exercise; the definition of transferable, which may exclude the managers you most want; and the operate period minimums that keep you paying markup after you are ready to own. BOT suits enterprises buying a 200 seat centre with procurement teams who negotiate those clauses. For a 20 seat pod, the same contract is a cage with good marketing.

🚀 Why EOR-seeded became the startup default

The EOR-seeded path inverts the sequence: hire the team first on an EOR's Indian entity, prove the centre works, then incorporate at the 20 to 30 head crossover and migrate. Speed is the visible advantage: days to first hire instead of months. The structural advantage is subtler: every rupee spent is linear and reversible, so the GCC decision stops being a leap and becomes a sequence of small, evidence based steps. The transfer at the end is a payroll migration on existing UANs, not a negotiation, because a flat fee EOR has no markup annuity to defend.

The honest caveat: past 40 or 50 committed seats with an India leader hired, starting DIY is defensible and eventually cheaper. The paths converge on the same destination; they differ on who carries the risk of being wrong early.

Here is the number the board asks for, built bottom up for a 20 person engineering pod: fifteen engineers across levels, two data roles, a designer, a product manager and an office manager, Bengaluru salaries, hybrid space.

💰 The two lines people miss

First, FX. A 3 to 5 percent bank spread on 700,000 dollars of annual payroll flow is 21,000 to 35,000 dollars, silently, every year. Settling at the RBI reference rate with zero spread deletes the line. It is the least glamorous saving in this table and the easiest. Second, recruitment: one time in the model, recurring in reality, because a growing pod hires every quarter. A 12 percent of CTC fee billed at day 90 keeps the incentive on retention; 25 percent billed at offer does not.

💸 The all-in per seat number

Blend the table and a well run 20 seat pod lands between 40,000 and 55,000 dollars per seat per year, all in, against 180,000 to 250,000 for the same seat in San Francisco or London. The arbitrage is real but it is the wrong headline. The teams that last are the ones built for capability at a better price, not the cheapest possible seat: pay the 60th to 75th percentile of the salary bands, fund the insurance, and the attrition line in your model gets to stay fictional.

Run your own numbers with the per employee calculator on our EOR services in India page; the table above is a pod level roll up of exactly that math.

The city decision gets over analysed and the first hires under analysed. In that order, then.

📇 The first three hires

Hire the anchor first: a staff level engineer or engineering manager who has worked with US or European teams and can hold your bar in interviews. Every subsequent hire calibrates to this person. Second, a senior individual contributor who ships visibly in month one, because the HQ sceptics are watching the first quarter. Third, the connective role, product manager or tech lead, who translates roadmap into the pod's rhythm. The office manager and the vanity director title can wait until head fifteen.

⚠️ The attrition myth, and the real version

India tech attrition headlines quote 20 percent plus, and badly run centres earn it. Well run pods, market pay, real ownership, a manager who is present, visible growth, run high single digits. Attrition is not weather; it is output. The inputs are in your control, and the cheapest one is scope: engineers stay where they own systems, and leave where they rent tickets.

⏰ Time zones, honestly

A GCC working with US teams sustains a four hour overlap comfortably; pretending to a full US shift burns the team inside two quarters. With Europe the overlap is generous by default. Write the expected overlap into every offer, and design the standups around the recorded and the asynchronous rather than someone's 11pm. The centres that treat time zones as a design constraint out-retain the ones that treat them as a negotiation.

A GCC's legal surface has three layers, and which ones you carry depends on the structure you chose in the build paths chapter.

🧾 Layer one: employment law

Whoever employs the people carries PF deposits by the 15th, TDS by the 7th, ESI where applicable, gratuity accruing at 4.81 percent of Basic plus DA, state registrations per working state, and exits settled inside the 48 hour full and final window. On the DIY path this is your subsidiary's job from day one. On the EOR-seeded path it is the EOR's, on the EOR's own registered entity, which is precisely what you are paying 149 dollars a month per head for. Either way the calendar is identical; the question is whose team runs it and whose name is on the penalty notice when the 15th is missed.

🧾 Layer two: transfer pricing

The moment an Indian entity does work for its foreign parent, the intercompany charge must be at arm's length, documented by an annual study, typically on a cost plus markup. Get it right and it is a compliance line item from the entity cost stack; get it wrong and assessments reopen years later. During an EOR-seeded phase this layer simply does not exist yet, because there is no intercompany transaction, one of the quieter advantages of sequencing the entity later.

⚠️ Layer three: permanent establishment

PE is the question of whether your India activity gives the Indian tax authority a claim on your global company's profits. The classic triggers: an office at your disposal, employees habitually concluding contracts in India, or a dependent agent acting mainly for you. A development pod doing internal R&D, paid at arm's length, with sales and contracting kept firmly outside India, sits in well charted territory. What creates real PE risk is drift: the pod lead starts negotiating with Indian clients, or a misclassified contractor team looks like a de facto branch. The defence is boring discipline: clean structure, documented charges, and sales authority kept where it belongs.

None of this should scare a founder off India; thousands of companies run this surface uneventfully. It should scare founders off improvising it. Every layer above has a professional owner available, from Big Four advisors on transfer pricing to an India native EOR on the employment layer. Buy the boring competence.

Here is the EOR-first sequence as it actually runs, with the gates where you decide to continue, pause, or change course. The dates assume a startup or mid market company starting from zero.

🔁 The migration, demystified

Because the employees sit on the EOR's Indian entity with proper PF from day one, migration to your subsidiary is a transfer, not a re-hire: new letters, PF riding on the same UANs, gratuity continuity written into the transfer terms, insurance novated. Teams experience a letterhead change. This is the single biggest structural difference from the contractor-first path, where formalisation means first time registration and awkward questions about the prior period, the trap covered in our contract staffing guide.

🤔 What if the experiment fails?

Then the structure fails gracefully, which is the point. A pod on an EOR winds down with notice periods, statutory settlements inside 48 hours, and 30 days notice to the EOR. No liquidator, no year of filings, no stranded entity. Compare that exit with the DIY path's one to two year wind down and the option value of sequencing becomes concrete. You are not paying an EOR fee for payroll processing; you are paying for the right to be wrong cheaply.

Gate discipline is the whole game. Write the gates down before Q1, with numbers, and let the India operation earn its structure. Centres built this way arrive at incorporation with evidence, a leader, and momentum. Centres built the other way arrive at month six with an entity, an empty office and a hiring plan still on slide four.

A GCC is not always the answer, and this chapter is the one the GCC consultants skip. The genuine decision is between four models, and the right one follows from two questions: is the work core, and is it permanent?

❌ When not to build one

Do not build a GCC to save money on work you do not care about; outsourcing does that with less management overhead. Do not build one to chase a competitor's press release; centres built as strategy theatre die in year two when the sponsoring executive moves on. And do not build one if headquarters is unwilling to send real work: a pod fed table scraps produces table scrap outcomes, attrition follows, and the retrospective blames India rather than the org chart. The failure mode is almost never the talent. It is the mandate.

✅ Where Versatile fits

One paragraph, then the verdict. We are an India native EOR, and the EOR-seeded path in this article is the thing we operationally run: 9 day shortlists, 5 day onboarding onto our own registered Indian entity, 149 dollars per employee per month with the first month free, payroll at the RBI reference rate with zero FX spread, and a designed, fee free hand off to your own entity at the crossover. The full machinery, including the statutory calendar and the cost calculator, is on our EOR services in India page.

If the work is core and the plan is permanent, build the GCC, and sequence it so every step is reversible until the evidence is in. That is the entire playbook, compressed.

The first wave of every GCC is engineering, because engineering is where the talent arbitrage is deepest and the output most measurable. The interesting question is the second and third wave, and the honest answer is that the map has widened far beyond what most boards assume.

🚀 The sequencing that works

Wave one: engineering plus QA, because the output is legible to HQ sceptics. Wave two: data and platform, which compound the first wave. Wave three: the operations functions, finance, support, people ops, once the site has management depth. Centres that invert this, starting with cost driven support functions, spend years fighting the low value site label and losing the engineers they later try to hire. The first function defines the brand of the centre internally, and internal brand is destiny for a GCC.

🤔 The AI wrinkle

The 2026 twist is AI teams. Model evaluation, data operations, fine tuning pipelines and applied ML engineering are being placed in India at striking rates, because the cost per experienced ML engineer in Bengaluru runs a third of the Bay Area price while the talent pool has compounded for a decade. For a startup, a five person applied AI pod is now one of the most common first GCC shapes we see, and it runs perfectly well on the EOR-seeded path: high salaries, small headcount, zero appetite for entity paperwork.

Four shifts define the current window, and together they explain why the playbook in this article looks different from the 2018 consulting deck.

📇 The entrants got smaller

The median new GCC used to be an enterprise programme. Now the growth is mid market and startup: 10 to 50 seat centres, often EOR-seeded, run by companies whose entire headcount is smaller than a classic captive's first floor. The enabling change is structural, the machinery for starting small: employer of record employment, managed offices by the seat, recruitment at 12 percent of CTC instead of retained search. The strategic change is talent: the engineers these companies want increasingly prefer product pods over services giants.

🧾 The law tightened, usefully

The four Labour Codes, operational from 21 November 2025, standardised the wage definition, Basic plus DA at a minimum half of CTC, recognised fixed term employment with pro rata gratuity, and pushed digital enforcement: PF defaults now surface in systems, not inspections. For well run centres this is good news, because compliance became more predictable and the cowboys became easier to spot. For anyone running grey structures, contractor fleets doing employee shaped work, the runway shortened sharply, a theme covered in depth in the contract staffing guide's legal chapter.

💰 Costs rose, the gap did not close

India tech salaries compound at 10 to 15 percent annually, and the arbitrage headline has narrowed for junior roles. But the comparison that matters, senior capability per dollar, still favours India by 3 to 4 times against US hubs, and the seniority of what is available has risen faster than its price. The 2026 GCC case is not built on cheap juniors; it is built on staff level engineers at a third of the Bay Area price with a decade of product experience behind them.

⏰ The window argument, without the hype

None of this says build now or miss out; India's talent pool is not running out. The genuine timing argument is internal: every quarter a committed India plan sits unexecuted costs the delta between your current burn on those roles and the India cost of the same output, and compounding starts when the pod does. If the two axes from the decision matrix already point at a GCC, the sequencing chapter means starting is a week's decision, not a year's. That is the actual change in 2026: the cost of finding out dropped to almost nothing.

Setting up the centre is a project. Running it well is a habit, and the habits that matter are visible within two quarters. The stalled GCC and the compounding GCC look identical on the org chart; the difference lives in four operating choices.

🚀 Own products, not tickets

The single strongest predictor of GCC health is what the teams own. A centre that owns products, services, or clearly bounded platforms compounds, because ownership creates the feedback loop that grows senior people. A centre that receives tickets from HQ stalls, because ticket queues cap growth at exactly the ceiling HQ sets. The test is blunt: can an engineer in the India centre ship something a customer notices without an approval crossing an ocean? If the answer is no in month six, it will still be no in year three, and your best hires will have noticed by month eight.

👥 Seniority first, then scale

The tempting sequence is to hire ten juniors fast and add leadership later. The sequence that works is the reverse: land one or two senior anchors, staff level engineers or a proven site lead, and let them pull their networks in behind them. Senior anchors set the bar for everyone hired after, and in a market where the strongest candidates evaluate the interviewer as hard as the interviewer evaluates them, the first five hires are your employer brand. This is also the strongest argument for the EOR-seeded path from the sequencing chapter: it lets you hire the expensive anchors first, on proper employment from day one, before any entity exists.

💰 The attrition math, honestly

India tech attrition runs high in services and lower in product centres, and the difference is not mysterious. People leave ticket queues and stay for ownership, leave opaque pay bands and stay for transparent ones, leave manager churn and stay for stable leadership. A well run product GCC holding attrition in the low teens is normal, not exceptional. Two mechanical points do heavy lifting: pay positioning at the 60th to 75th percentile of the city band costs far less than the recruitment and ramp cost of the churn it prevents, and a real promotion path inside the centre, not one that requires relocation, keeps your seniors from becoming somebody else's anchors.

🤔 The cadence that keeps it honest

The compounding centres share a boring rhythm: a weekly leadership sync that treats India as a peer room rather than a status call, quarterly travel in both directions, and one metric reviewed monthly that measures output, shipped work, not activity. The stalled centres share the opposite: HQ visits once a year, India leadership hears decisions second hand, and the monthly review counts hours. None of this costs much. All of it decides whether the cost table from the earlier chapter buys you a real product organisation or an expensive room of people waiting for instructions.

🤔 Is there a minimum viable GCC size?

Smaller than the consultants say. A pod of five with a strong anchor engineer ships meaningful product; below that you have remote employees, which is also fine, just not a centre. The floor that matters is management attention: one accountable leader who treats the pod as their team rather than a vendor relationship. Headcount grows from evidence; attention has to be there on day one.

🤔 Is IP safe in an India centre?

India's IP statutes are TRIPS aligned and employment contracts carry enforceable assignment and confidentiality provisions; every major technology company on earth runs core engineering here, which is the market's own due diligence verdict. The practical risks are operational, not statutory: access management, device policy, and the difference between employees under Indian employment contracts and loosely papered contractors. Employment through a proper entity, yours or an EOR's, is the strong version of the paper.

🤔 How do we stop the GCC becoming a separate culture?

Culture drift is a distance problem, not an India problem. The mitigations that work are unglamorous: the pod attends the same standups and demos as HQ, ships from the same repos with the same review standards, gets the same information in the same all hands, and sees its members promoted into org wide roles on the same ladder. Fly the anchor to HQ in quarter one and an HQ lead to India in quarter two. Culture follows the work and the ladder; posters follow neither.

🤔 What goes in the board deck?

Four numbers a quarter: all in cost per seat against the table in the cost chapter, attrition against a high single digit target, shipped roadmap items attributable to the pod, and time to fill against a 9 day shortlist standard. Plus one structural line: heads on EOR versus own entity, with the crossover decision date. Boards do not need the incorporation saga; they need evidence the capability is compounding and the structure is deliberate.

🤔 EOR now, entity later: does the switch disrupt the team?

It is a letterhead change when run properly: same UAN, PF continuity, gratuity terms carried over, insurance novated inside the same cycle. The disruption risk sits in vendors with exit fees or transfer clauses priced per head, which is why those two lines on any rate card deserve more scrutiny than the headline fee. Ours are zero, by design, because the model only stays honest if leaving is cheap.

🤔 What does the first quarter actually cost us in leadership time?

More than the invoices, and it is worth budgeting honestly. Expect your CTO or VP Engineering to spend a real fraction of their week on India during the first quarter: interviews for the anchor hires, two or three working sessions a week with the new pod, and one trip. Teams that budget this time get a centre that runs itself by month six. Teams that treat India as a procurement exercise, sign the contract and delegate the attention, get the stalled version from the operating chapter and blame the model. The money is the easy part of a GCC. The attention is the actual investment, and it is front loaded: heavy for one quarter, light thereafter.

⭐ The position, stated once. Build a GCC when the work is core and the plan is permanent, and build it for capability at a better price rather than the cheapest seat in the market. Sequence it so being wrong is cheap: seed the pod on an EOR in days, hold quarterly gates with numbers, incorporate at the 20 to 30 head crossover once an India leader exists, and migrate on one payroll cycle. Skip the BOT contract unless procurement can negotiate transfer clauses, and skip the GCC entirely for work you would not miss.

Judge the market's vendors on the same four numbers as always: days to shortlist, days to onboard, the shape of the fee over three years, and the cost of leaving. Capability centres are decade assets; buy structures that expect you to outgrow them.

I run Versatile, an India native EOR, and seeding GCCs is our bread and butter: your first pod employed on our own registered Indian entity inside a week, PF, ESI, TDS and gratuity run in house across 28 states, 149 dollars per employee per month, first month free, no setup or exit fees, zero FX spread at the RBI reference rate, and a hand off to your own entity that we treat as a graduation rather than a churn event. If a GCC is in your next two quarters, start with the EOR machinery or use the form on this page and I will reply personally.

One last filter for the board deck: if the plan cannot name the first two anchor hires and the product they will own, it is not a GCC plan yet, it is a cost model. Get those two names and that one product right and the rest of this article is just sequencing. Everything else, the city, the office, the legal wrapper, can change later without breaking anything. Those two choices cannot.