Everything that decides whether an India GCC works

Developments log

New entries are added as GCC policy, cost, tax and market developments land. Corrections to the body of the playbook are flagged here too, so nothing changes silently.

What a GCC actually is

A Global Capability Centre is a wholly-owned offshore entity of a parent company that employs its own people to do the parent's work.

The defining test is employment and accountability: the engineer in Bengaluru is on the parent group's payroll, carries the parent's email domain, and sits in the parent's org chart. Everything else — size, charter, location, maturity — varies.

That single test is what separates a GCC from outsourcing. In outsourcing, the vendor employs the people, owns the delivery risk, prices the output, and keeps the margin. In a GCC, the parent employs the people, owns the delivery risk, pays cost plus a regulated markup to itself, and keeps the capability. Outsourcing buys capacity; a GCC builds an asset.

The vocabulary, used correctly

The four things a GCC is built to do

Most centres are justified on one of these, then drift across the others as they mature:

1. Cost — do the same work for 40–60% less total operating cost.
2. Talent access — reach a skill pool that does not exist, or is unaffordable, at headquarters.
3. Scale and speed — hire 200 engineers in a year, which the home market cannot absorb.
4. Control and IP — keep proprietary product, data or domain knowledge inside the group rather than with a vendor.

The market, in validated numbers

India hosts roughly 1,700+ GCCs across nearly 3,000 units, employing about 1.9 million people, growing at roughly 9.8% CAGR.

The three shifts that matter commercially

1. The buyer has changed. The growth is no longer Fortune 500 first-centres — it is mid-market companies with $100m–$1bn revenue. 480+ mid-market GCCs now operate in India and 35% were established in the last two years. These buyers have no India playbook, no internal GCC team, and no patience for a 12-month build.

2. Mid-market centres mature faster, not slower. Despite operating at roughly 40% the scale of large-enterprise GCCs, mid-market centres show 1.3x higher presence as transformation hubs and advance along the maturity curve 1.2x faster. A smaller parent cannot afford a 500-person execution-only centre, so the India team gets real ownership earlier. India now holds 47% of global product management talent for mid-market GCCs.

3. Geography is decentralising under policy pressure. Budget 2025 announced a National Framework for GCCs aimed at tier-2 cities, and Karnataka, UP, Telangana and Gujarat have all published competing GCC policies with cash incentives. Tier-2 is the fastest-growing segment at 23% YoY.

How to use these numbers

The aggregate numbers are credible for positioning and keynotes; they are not a business case. No client decision should rest on "the market is $100bn." The numbers that move a deal are cost per FTE, payback month, and attrition-adjusted run rate. Lead with market data, close with unit economics.

Greenfield vs brownfield, defined properly

Most writing on this borrows the manufacturing definition — new land versus existing plant — and applies it to GCCs, where it does not fit.

In a GCC context the distinction is not about buildings. It is about whether you are starting from zero organisational footprint or absorbing one that already exists.

Greenfield: no prior legal entity, no employees, no leases, no vendor relationships, no India delivery history. You incorporate, lease, hire and build the culture from the first employee onward. Full design freedom, full timeline, full cost.

Brownfield: an India footprint already exists in some form and you are converting, absorbing or restructuring it. There are four distinct brownfield situations, and conflating them is the most common analytical error in this field.

The four kinds of brownfield

Cost, timeline and risk compared

The judgement call

Greenfield is slower and costs more, and almost always produces a better centre. Brownfield is faster and cheaper on paper, and the savings are routinely consumed by remediation — fixing comp structures, correcting misclassified contractors, clearing statutory arrears, replacing inherited talent.

The five entry models

Most advisory conversations collapse these into two — build or buy — which loses the two models that actually fit mid-market buyers.

How each model actually costs out

• Direct captive: cheapest steady state, most expensive start. No partner margin on the run rate. Right when the three-year commitment is genuinely certain.
• BOT: a build fee, a vendor margin during the operate phase, then a transfer fee. Compresses go-live to as little as six months against 12–24 for an unassisted greenfield. The transfer price is the negotiation — fix it in the original contract, never leave it to "fair market value at the time."
• EOR-first: $99–$399/employee/month with India specialists, $499–$699+ with global platforms, on top of salary plus 20–25% statutory. Setup is $0–$1,000 against $15,000–$25,000+ for an entity. Crossover to an owned entity is around 12–40 employees, or roughly $500,000 annual offshore spend. An own entity adds $2,000–$3,500/month of standing overhead per state.
• Managed GCC: all-in managed workspace in Bengaluru runs roughly ₹8,000–₹15,000 per seat per month mid-market, up to ₹24,000 at premium finish. You avoid the fit-out capex entirely and keep the employment relationship, which is the part that matters.
• Carve-out: budget remediation at 15–25% of first-year people cost for comp correction, statutory clean-up and replacement hiring.

The decision matrix

The decision gate: should they build one at all?

A GCC is a fixed-cost, long-duration commitment dressed up as a cost-saving initiative.

Most failed centres were approved by people who had not tested these six gates. Running these honestly — including concluding that the answer is no — is what separates a real assessment from a sales process.

The six qualifying gates

1. Scale. Is there a credible path to 25+ FTE within 18 months? Below that, fixed overhead is not amortised and per-FTE cost stays near home-market cost. Under 25 people, the answer is EOR.
2. Duration. Committed for three years minimum? Setup cost is recovered over 18–36 months. Ask directly: what would have to happen for you to shut this down in year two?
3. Work-definability. Can the work be specified well enough to be executed 8,000 km away? Work that depends on daily corridor conversations with a founder does not transfer.
4. Sponsor. Is there one named executive at HQ whose own objectives depend on this succeeding? Not a steering committee. This is the strongest single predictor of success, and the most common gap.
5. Headroom. Enough work for three years, or hiring against a hope? A centre that runs out of mandate at month 14 enters the attrition spiral immediately.
6. Tolerance. Can the organisation absorb 15–25% attrition, a 3–6 month leadership search, and a first-year productivity dip? If the CFO expects savings in quarter two, the project will be killed before it works.

The disqualifiers — say no to these

• The driver is purely a cost target from a PE sponsor, with no capability intent. These get cut in the next portfolio review.
• Headcount under 15 with no growth plan. Use an EOR instead.
• The parent is in distress, mid-sale, or between CEOs. Entity decisions do not survive a change of sponsor.
• The work is a single client contract that could be lost. The centre dies with the contract.
• HQ leadership privately believes offshore work is lower quality. Culture beats business case.

When outsourcing is the honest answer

Recommend a vendor when the work is non-core and unlikely to become core; when volume is genuinely variable and needs to flex down as well as up; when scale never exceeds 20 people; or when the skill is so niche that no sustainable India bench exists. Saying this out loud, early, is what makes the rest of the assessment credible.

Location strategy

City choice moves total cost by 25–50% and attrition by 10–15 percentage points. It is routinely decided on the basis of where the CEO has visited.

Tier-1 cities compared

Fit-out figures are Cushman & Wakefield's 2026 city table for a collaborative hybrid workplace specification — one finish standard, not a basic-to-premium range — corroborated within 1–2% by JLL (India medium-spec average ₹5,788; Mumbai ₹6,588). Two cautions: C&W publishes a "Delhi" figure, so applying it to Gurugram or Noida is an extrapolation the table does not make; and C&W covers eight metros only, so every tier-2 fit-out number below is inferred, not sourced. Rent and attrition columns are directional — the Knight Frank city rent table failed verification.

Tier-2 cities

Tier-2 markets deliver 25–50% lower real estate, talent and operating costs, and attrition of 10–12% against 18–22% in metros, largely because there are fewer competing employers to poach into. The working rule: 70–85% of tier-1 talent quality at 70–85% of tier-1 cost. Mid-level engineers in Coimbatore, Jaipur and Kochi command 25–35% less than Bengaluru equivalents for comparable technical depth.

The multi-hub model

The emerging default above 300 FTE is not one city — it is a tier-1 anchor plus a tier-2 scale site. The anchor holds leadership, architecture and product management; the tier-2 site holds engineering pods, QA, support and operations where retention matters more than market depth. Blended cost lands 15–25% below a single tier-1 site.

How to actually run a site selection

Score candidate cities on seven weighted criteria rather than arguing about them:

1. Talent availability at your specific skill — not "IT talent." Count live job postings and competitor headcount in that city.
2. Salary index for those roles against Bengaluru.
3. Competitive intensity — how many employers will bid for your people. This drives attrition more than anything else.
4. Real estate cost and Grade-A availability including vacancy rate.
5. State incentive value, net of the conditions attached.
6. Connectivity — direct or one-stop flights from HQ. Executive visit frequency correlates with centre success.
7. Leadership attractability — can you persuade a ₹1.5–3 crore GCC head to live there? This disqualifies more tier-2 cities than any cost factor.

Weight talent availability and competitive intensity highest. Cost differences between shortlisted cities are usually 10–20%; a wrong talent call costs 40%+ in attrition and replacement.

The build sequence

A greenfield captive build reaches go-live in 4–6 months; a managed or EOR-first build in 6–12 weeks.

The difference is not effort — it is how much of the critical path you own. The two tracks that must start in week one are the ones with the longest lead times and no shortcut: the leadership search and the entity incorporation.

The critical path runs through leadership, not through the entity. Incorporation takes 8–12 weeks of largely parallel paperwork; finding a GCC head takes 3–6 months of search. Start both on day one and the entity is ready before the leader is.

What each track actually contains

Mandate and charter (weeks 1–3). Write the charter: what the centre does, what it explicitly does not do, the headcount curve, and the named HQ sponsor. One page. If this cannot be written in week three, the project is not ready.

Site selection (weeks 2–5). Score cities on the seven criteria. Visit the shortlist. Do not sign anything.

Leadership search (weeks 1–22). Launch immediately. Budget 3–6 months and expect one failed offer. The first senior hires come through this person's network, so every week of delay compounds.

Entity incorporation (weeks 3–14). SPICe+ filing, resident director, PAN/TAN, GST, bank account, FIRMS entity master, then payroll registrations. The usual delay is apostilled parent documents — start collecting them in week one.

Workspace (weeks 5–16). Managed office compresses this to 3–4 weeks and removes the fit-out capex. A conventional lease plus fit-out adds 3–6 months and is the single largest source of slippage.

Tax, TP and compliance (weeks 9–18). Intercompany MSA, cost allocation policy, IP assignment, transfer pricing position, compliance calendar, payroll provider. All live before the first invoice to the parent.

IT and security (weeks 11–20). Network, endpoints, identity, access controls, DLP, VPN or zero-trust. For regulated parents this track, not the entity, becomes the critical path — scope it in week one.

Hiring wave 1 (weeks 9–18). Senior roles first. Note the deliberate overlap: hiring begins before the workspace is finished, which is why interim or managed space matters.

The gates

1. Week 5 — charter and city confirmed. No charter, no lease, no incorporation.
2. Week 14 — entity live and leader signed. If the leader has not signed, pause hiring rather than letting an HQ manager build the first cohort.
3. Week 18 — go-live. Entity, workspace, IT, compliance and the first cohort all in place.
4. Week 30 — ramp checkpoint. At 60–70% of planned headcount, with attrition and productivity measured, commit to the long lease. Not before.

Legal and entity architecture

For a GCC there is effectively one right answer: a wholly-owned subsidiary structured as a private limited company.

The alternatives exist and almost all are wrong for this use case. Knowing why is what makes you credible with a client's general counsel.

The FDI position

IT, ITES, software development and most GCC activity sit under the automatic route — 100% foreign ownership, no prior approval. The government route applies to restricted sectors such as defence, telecom and media. Confirm the cap against the activity, not the industry: a media company's engineering GCC is usually automatic-route IT, but one doing broadcast operations may not be.

Incorporation mechanics

Incorporation runs through SPICe+ on the MCA portal and realistically takes 8–12 weeks for a foreign parent, with the delay almost always in apostilled parent documents, not the filing.

• Minimum two directors, at least one resident in India (182+ days). The most common cause of delay — identify them in week one.
• Minimum one shareholder; a foreign company may hold 100%.
• DINs and digital signature certificates for all directors.
• A registered office address in India from day one.

The registration stack

The ongoing FEMA filings

These get missed and later surface as penalties during diligence:

• FC-GPR — on every share allotment to the parent, within 30 days.
• FC-TRS — on any share transfer between residents and non-residents.
• FLA return — annually by 15 July, mandatory for any company with FDI, even dormant.
• APR — where applicable to overseas investment structures.

The intercompany paper that must exist

A GCC with no service agreement is a transfer pricing adjustment waiting to happen. Before the first invoice:

1. Master Services Agreement specifying scope, service levels, cost base and markup methodology.
2. Cost allocation policy — what sits in the cost base and what is passed through.
3. IP assignment clauses confirming all work product vests in the parent. Retrofitting this across 200 employment contracts is expensive.
4. Secondment agreements for HQ staff working in India and vice versa — a permanent-establishment risk if left informal.

One-time setup

Total one-time setup runs $500,000 to $3 million (roughly ₹4.75–28.5 crore), depending on scale and finish.

The range is wide because fit-out and IT dominate, and both scale with headcount and ambition rather than with the entity work. USD–INR converted at approximately ₹88/$ — restate at the live rate for any client proposal, since a 5% currency move changes a 250-FTE business case by more than most line items.

Line-item build-up

Setup cost by team size

Where the money actually goes wrong

• Fit-out is the single biggest variance. Indian design-build contractors quote ₹800–2,500/sq ft for basic-to-mid finish, against C&W's 2026 collaborative-hybrid benchmark of ₹5,847–6,567/sq ft — a 3–4x span on the largest capex line. Treat these as two different measurement bases rather than two ends of one range, and establish which basis a quote is on before comparing it to anything.
• Seat density drives everything. At 80 sq ft/seat versus 120 sq ft/seat, a 100-person centre needs 8,000 versus 12,000 sq ft — a 50% difference in both rent and fit-out. A seat ratio below 1:1 is normal under hybrid working, but no independent source supports a specific number — the commonly cited 1.5 headcount per seat benchmark failed verification in October 2026. Derive the ratio from the client's own observed attendance.
• Security deposits are forgotten in the business case, then surface as a cash-flow surprise at lease signing.
• Recruitment fees are systematically underestimated. At 10% average on a 100-person first-year ramp with 20% attrition replacement, that is 120 hires, not 100.
• Managed office eliminates roughly $95k–$540k of this table in exchange for a higher per-seat run rate. For anything under 75 FTE or any uncertain charter, that trade is usually correct.

The steady-state run rate

People cost is 70–80% of total operating cost per FTE. Everything else is rounding.

Which is why location and comp strategy matter more than any procurement saving on rent or hardware.

Salary bands — India GCC, 2026

Annual CTC in ₹ lakhs, Bengaluru-benchmarked. Apply the city index: Hyderabad 0.88–0.92, Pune/Chennai 0.75–0.85, tier-2 0.65–0.75.

GCCs pay 25–40% above Indian IT services firms at mid-senior level — budget to that premium or lose every competitive offer.

Statutory loading — the exact build-up

Employer statutory cost adds 15–22% on top of base salary (some providers quote 20–25% including gratuity accrual and statutory bonus):

Annual operating cost — all line items

The hidden costs that break business cases

1. Replacement hiring at scale. 20% attrition on 250 FTE is 50 hires a year, at recruitment fees plus 3–6 months of lost productivity each. Usually absent from the original model entirely.
2. Lease escalation. 5–7% annually on a nine-year lease compounds to roughly 1.7x by year nine. Model it.
3. Salary inflation outrunning the arbitrage. India increments ran 9.1% all-India and 9.3% in GCCs specifically for 2026 (Aon), with GCC pay inflating at or slightly above the market rather than below it. Indian IT services firms increment at only 6.6%, a persistent 2.7 percentage-point gap that makes GCCs the pay-setter. Note what cannot be said: without a validated US increment figure the annual erosion rate of the arbitrage cannot be computed, so do not quote one.
4. The transfer pricing markup. A 15.5–18% markup on the full cost base is a real cost to the parent and is frequently omitted from the savings calculation.
5. Night-shift and statutory transport obligations for teams covering US hours.
6. Dual-running costs during transition. If the GCC is absorbing work from HQ or a vendor, you pay twice for 3–9 months.
7. The counter-offer market. Notice periods of 60–90 days are standard; buyouts and retention bonuses to close senior offers are now routine in Bengaluru.

Per-FTE economics and the arbitrage math

The number that decides a GCC is fully loaded annual cost per FTE. The number that gets it approved is the month the cumulative saving crosses the cumulative investment.

Year-one total cost by team size

Note the shape of the curve — per-FTE cost falls only about 13% going from 25 to 250 people. Scale economics in a GCC are real but modest. The large savings come from location and role mix, not from size.

Steady-state per-FTE cost

Once setup is amortised, steady-state run rate is $25,000–$80,000 per engineer per year, roughly 40–60% below a comparable US build. The spread is driven almost entirely by seniority mix and city.

What savings are real, and what are illusory

Real: base salary differential, especially at junior and mid levels; real estate cost per seat; 24-hour coverage without US overtime premiums; retained institutional knowledge versus vendor churn.

Illusory or overstated:

• The savings number quoted before the TP markup. A $50k India FTE costs the parent ~$57.75k after a 15.5% markup.
• Productivity parity assumed from month one. A realistic ramp is 40% productive in months 1–3, 70% in months 4–6, 90%+ from month 7 — roughly half a year of partial output per hire, repeated on every replacement.
• Senior role savings. At staff-engineer and director level the gap has compressed to 50–65%, and for scarce AI/ML talent it can be under 50%.
• Management overhead at HQ. Someone at headquarters now spends 20–40% of their time on India coordination. Real, but invisible in the India P&L.
• Savings on a 25-person centre. Fixed overhead swamps the arbitrage.

The payback model

Build the case as a cumulative cash-flow crossover, not an annual saving percentage:

1. Baseline: what the same work costs today at HQ or with a vendor, fully loaded.
2. India cost: year-one total including setup, plus the TP markup.
3. Ramp adjustment: discount India output by the productivity curve for the first two quarters.
4. Crossover month: typically month 14–24 for a 100-FTE greenfield; month 6–10 for EOR-first or managed.
5. Three-year NPV at the client's hurdle rate, with India salary inflation at ~9.2% and lease escalation at 5–7%.

Case study: Northvale Clinical Systems

The company and the trigger

Northvale Clinical Systems is a US-headquartered clinical trial software company — $180m revenue, 740 employees, PE-backed, Boston HQ. Three pressures arrived at once: engineering hiring in Boston was taking 94 days at an average loaded cost of $205,000 per engineer; their Pune outsourcing vendor had raised rates 11% and lost two key people on the account; and the PE sponsor wanted 300 basis points of margin improvement before the next funding round.

The mandate: build a 150-FTE engineering and clinical-operations centre in India within 18 months.

The charter

Model and location

EOR-first, converting to own entity at month 5, in a managed office. The PE sponsor would not approve $2m of capex on an unproven thesis, but would approve an operating expense that could be stopped. Hiring started in week 3 via EOR while incorporation ran in parallel.

Hyderabad. Bengaluru offered the deepest product pool but 25% higher rent and 25% attrition. Hyderabad gave a strong life-sciences technology cluster, rent at ₹75/sq ft, salary index 0.90, and attrition near 18%. Pune was eliminated to avoid non-solicit conflict with the incumbent vendor.

Ramp plan

Three-year financial model (USD)

Cumulative net saving crosses cumulative investment in month 11. Three-year net saving: approximately $19.6m. Steady-state unit cost $63,870 per FTE against a $131,000 blended baseline — a 51% reduction, squarely inside the validated 40–60% band.

What Northvale got wrong

1. They hired the GCC head at month 4, not month 1. The first 15 hires were made by a Boston engineering director over video calls. Six of those 15 left within a year — a 40% first-cohort attrition rate against 18% for everyone hired after the GCC head joined.
2. They leased 120 seats for 150 people and still ran out of space. Hybrid policy drifted to four days in office once teams formed, because new teams need proximity. The seat ratio assumption was made from HQ's hybrid norms, not India's.
3. They omitted the TP markup from the board paper. The first board review showed costs 15.5% above plan with no explanation, which cost the sponsor credibility for two quarters.
4. They under-budgeted replacement hiring by half. The model assumed 15% attrition; actual was 21% in year two, concentrated in the 6–10 year engineering band — exactly where the market data says it concentrates.

Team structure and org design

The structural question that decides whether a GCC becomes a capability centre or stays a delivery shop is not the org chart — it is whether the India team reports into work or into a function.

A pod that owns a product area and reports to a product leader builds capability. A pool of engineers assigned to tickets from HQ does not, no matter how it is drawn.

The GCC head — the single most important hire

Hire this person first, before the first engineer. Northvale's 40% first-cohort attrition is what happens otherwise, and it is the most reliably repeated mistake in the field.

• 15–25 years' experience, with prior experience building a centre from under 50 to over 200 people — not running a mature 2,000-person site. Different jobs.
• Credible to HQ engineering leadership on technical substance, not just operations.
• Deep local network for senior hiring — the first 20 senior hires come mostly through them.
• Willing to do operational work for the first year. At 40 FTE there is no HR team and no one else to fix the internet.
• Reports to a business leader at HQ, not to HR or to a shared-services function. Reporting line signals charter more than any mandate document.

Budget ₹140–450 lakhs CTC depending on scale and charter, plus a 25–33% search fee. Expect a 3–6 month search. Start it in week one.

Structure by stage

The ratios to plan against

Solid line vs dotted line — the real decision

1. Functional solid line to HQ. The India engineer reports to an HQ manager; the GCC head owns only site operations. Fast to set up, and the known route to the career-ceiling problem. The GCC head becomes a facilities manager with a big title.
2. Dual reporting. Solid line to India management, dotted to HQ function. The workable middle, most common at 100–300 FTE. Requires genuine agreement on who sets priorities and who writes the review.
3. India-owned pods with HQ as stakeholder. The India leader owns the product area and outcome; HQ specifies what, India decides how. The only structure that produces a transformation hub.

The first ten hires, in order

1. GCC head
2. Senior engineering manager / tech lead — hires and sets the bar for the team
3. HR / talent acquisition lead — before the hiring volume arrives
4. First senior engineer, ideally someone the GCC head has worked with
5. Second senior engineer
6. Finance and compliance manager (fractional is fine under 50 FTE)
7. IT / infrastructure engineer
8. Second tech lead or first functional specialist
9. First mid-level engineers (9–10)

Seniority first, then volume. Every GCC that hires 20 juniors before it has two senior engineers spends the following year fixing it.

Total rewards and equity

Seventy per cent of senior attrition is about progression and ownership rather than pay. The instrument that actually confers ownership is equity, and most mid-market parents never extend it to India.

Section 10 covers what to pay. This section covers the rest of the package, which is where the retention argument is either won or quietly lost. A salary band is a floor. It is not a reason to stay.

The benefits stack — mandatory versus market-expected

Equity: the real ownership instrument

A parent can extend its own stock plan to Indian employees of its subsidiary, and whether it does is one of the sharpest signals a GCC sends about how it regards the India team. An engineer choosing between a product company offering RSUs and a GCC offering none reads that correctly as a statement about status.

The mechanics differ materially between the two instruments, and the difference matters to the employee’s cash flow, not just the tax bill.

The FEMA and reporting side

Cross-border grants fall under the Overseas Investment Rules, 2022 and are treated as Overseas Portfolio Investment. Two obligations follow, and both sit with the Indian entity rather than the employee:

• Form OPI — the Indian employer reports the overseas portfolio investment within 60 days of the end of the relevant half-year.
• Annual ESOP return to the RBI covering shares held by employees under the plan.

Where the foreign parent withholds tax in its home jurisdiction, the employee claims relief under the applicable Double Taxation Avoidance Agreement. Mechanics per Legal 500 on cross-border ESOPs for GCCs and Treelife on foreign-parent ESOPs.

The design question worth asking first

Before any of the mechanics: does the parent intend to extend its plan to India at all, and at what level? There are three honest answers, and they produce three different centres.

1. No equity for India. Common in mid-market and PE-backed parents. Defensible on cost, but it caps the centre at execution work, because the people who build products expect to own some of the upside. The career-ceiling problem in section 18 follows from it.
2. Equity for leadership only. The usual compromise. Retains the GCC head and a handful of directors; does nothing for the 6–10 year engineering band where attrition actually concentrates.
3. Equity on the same terms as the home market, by level. Rare, expensive, and the single clearest signal that India is part of the company rather than a supplier to it.

This is a structuring decision taken at the start, because retrofitting equity into an established comp structure is far harder than granting it from the first cohort.

Compliance and governance

India compliance is not difficult, it is relentless — a high volume of small, dated obligations where the penalty for missing one is disproportionate to its size.

The failure mode is never complexity; it is someone forgetting. Build the calendar before the first employee.

The recurring compliance calendar

The labour law stack

The four Labour Codes took effect on 21 November 2025, consolidating 29 earlier statutes. State-level rules are still being notified in places, so confirm the position in your state, but the Codes are now the operative framework rather than a pending reform. The obligations that bind a GCC:

• Shops and Establishments Act (state) — hours, leave, holidays, women working night shifts, mandatory transport and security for late shifts.
• Payment of Gratuity Act — 15 days' wages per year of service, payable after five years. Accrue from day one.
• Maternity Benefit Act — 26 weeks paid leave. Mandatory crèche at 50+ employees. Frequently missed.
• POSH Act — Internal Committee mandatory at 10+ employees, with an external member, plus annual report filing. The single most commonly neglected item in new GCCs.
• Contract Labour Act for housekeeping, security and transport vendors — principal-employer liability sits with you if the vendor defaults.
• Rights of Persons with Disabilities Act — accessibility obligations on premises.

Data protection — the DPDP Act

The Digital Personal Data Protection Act, 2023 is the governing framework, with rules phased in: consent and notice, security safeguards, breach notification to the Data Protection Board, purpose limitation, and a Data Protection Officer where the entity is a Significant Data Fiduciary.

For GCCs the practical question is almost always what the parent's data obligations impose on India, not what Indian law requires. A GCC handling EU personal data needs GDPR-grade controls and Standard Contractual Clauses between parent and subsidiary; one handling US healthcare data needs HIPAA Business Associate Agreements; one in financial services inherits the parent's regulator. Map the parent's regime first, then layer DPDP on top. This is routinely discovered six months late, when remediating access controls across a live centre is expensive.

SEZ and STPI — the honest position

SEZ: the headline income tax holiday under section 10AA has sunset for new units — the deadline for commencing operations passed on 31 March 2020. New GCCs get no income tax holiday from an SEZ. What remains is duty-free imports and GST treatment, against significant procedural burden and location constraints. For most new GCCs, SEZ is no longer worth it. Say so plainly; a surprising number of advisors still pitch the lapsed benefit.

STPI: still useful. Duty-free capital goods import, simplified export documentation through Softex filings, single-window support, low compliance overhead. Worth doing for hardware-heavy centres; marginal for pure cloud operations.

Governance structure

• Board: minimum two directors, one resident in India. Keep at least one HQ executive on the board — it forces engagement.
• Statutory registers and minutes maintained properly from incorporation. Examined in every diligence; gaps are expensive to fix retrospectively.
• Operating governance, separate from statutory and more important day to day: a monthly operating review with the HQ sponsor, a quarterly business review with HQ leadership, and an annual charter review that explicitly asks whether the mandate should expand. Without that last one, the charter never grows.
• Delegation of authority matrix — what the GCC head can sign without HQ approval. Set at incorporation. An India leader who needs sign-off for a ₹50,000 purchase cannot run a centre.

Transfer pricing — the economics nobody models

Transfer pricing is where GCC business cases quietly lose 15–18% of their claimed savings.

Understanding it properly is one of the clearest ways to be the most credible person in a GCC conversation, because most recruiters and many advisors cannot explain it.

The mechanic

A GCC does work for its parent. Indian tax law requires that transaction to be priced at arm's length. For a captive service centre the accepted method is cost plus: the India entity charges the parent its total operating cost plus a markup, and pays Indian corporate tax on that markup as its profit.

So a centre with a $6m cost base and a 15.5% markup invoices the parent $6.93m. The $930,000 of profit is taxed in India. The parent's true cost is $6.93m, not $6m. Any savings calculation that uses the cost base rather than the invoiced amount overstates the saving by the full markup.

Safe harbour margins

The 2026–27 reform is materially favourable and under-discussed: the margin drops from 17–18% to 15.5%, the eligibility ceiling rises from ₹300 crore to ₹2,000 crore, and four previously separate service categories are unified. A ₹2,000 crore ceiling covers virtually every mid-market GCC and most large ones.

Safe harbour, benchmarking study, or APA

What actually gets challenged

India's transfer pricing officers have moved toward larger adjustments, and GCC markups are an explicit focus area alongside intra-group service charges, financial guarantees and tax-holiday units.

1. Cost base composition. Stock compensation, HQ-allocated costs and one-time setup costs are all contested.
2. Characterisation of the entity. The authority may argue the India centre is not a low-risk contract service provider but a genuine entrepreneur bearing risk and owning IP — justifying a far higher profit allocation to India. The more capability the GCC genuinely owns, the weaker the cost-plus characterisation becomes. The single most important forward-looking TP issue in the industry, and almost nobody raises it.
3. Comparables selection in a benchmarking study.
4. Marketing intangibles where the India entity's work builds parent brand value.
5. Loss-making or low-margin years — a captive is expected to earn a stable return regardless of parent performance.

Documentation that must exist

• Form 3CEB — accountant's report on international transactions, due 31 October.
• Local file — TP study with functional and economic analysis, maintained contemporaneously. Required where international transactions exceed ₹1 crore.
• Master file (Form 3CEAA) where group thresholds are met.
• Country-by-Country report above the consolidated revenue threshold.
• Intercompany agreement matching actual conduct. Where the paper and the practice diverge, the practice governs — and the paper becomes evidence against you.

State incentives and policy

Since 2024, Indian states have been competing openly for GCCs with published policies and cash incentives. This is new, it is live, and most mid-market buyers have no idea it exists.

The national framework

Budget 2025 announced a National Framework for Global Capability Centres, guiding states to promote GCCs in emerging tier-2 cities, focused on skill development, infrastructure and AI-focused centres of excellence. Important caveat: unlike state policies, the national framework offers no capital subsidies or tax breaks. It is coordination. The money is at state level.

How to actually monetise an incentive

The gap between a published incentive and cash in the bank is where most of these schemes die:

1. Apply before you commit. Almost every scheme requires approval before the investment is made or the lease signed. Retrospective applications are routinely rejected — this alone disqualifies most companies that discover incentives after choosing a city.
2. Read the employment commitment. Incentives are tied to minimum headcount sustained 3–5 years, with clawback if you fall short. If the ramp is uncertain, a smaller incentive with a lower commitment is worth more.
3. Opex reimbursement is claimed, not granted. Requires documented quarterly or annual claims with evidence. Unclaimed incentives are the norm, not the exception.
4. Capital subsidies are disbursed in tranches against verified milestones, often with a 12–24 month lag. Do not model them as year-one cash.
5. Negotiate directly. Above roughly 300 FTE, published policy is a floor. States negotiate bespoke packages, and a committed anchor tenant has real leverage.

Failure modes

GCCs rarely fail loudly. They fail by stalling — 60 people, a transactional charter, 25% attrition, delivering enough not to be shut down and never enough to matter.

1. The attrition spiral

Voluntary attrition in Indian GCCs averages 16–22%, rising to 25–30% for AI/ML and senior engineering roles, and 51% of GCCs name talent retention as their top concern.

Mitigation: give India pods end-to-end ownership of a product area; create a technical career ladder that reaches principal and distinguished engineer with real authority; put at least one India-based leader in a global role within 24 months; and measure the proportion of India roles that are global rather than site-specific.

2. The career-ceiling problem

Specific to the GCC model. India teams report into HQ functions that set strategy, so India executes and does not decide. Engineers with 6–10 years' experience read this correctly as a ceiling and leave — exactly the band where replacement is hardest and most expensive.

Mitigation: the ownership structures in section 13. There is no workaround that preserves execution-only reporting.

3. Mandate creep and charter starvation

Two opposite failures with the same cause — no charter discipline. Creep: the centre accumulates whatever HQ wants to offload and develops no deep capability in anything. Starvation: HQ promises scope that never arrives, the centre runs out of meaningful work by month 14, and good people leave.

Mitigation: a written charter naming what the centre does and explicitly what it does not; an annual charter review with the sponsor; and a rule that new scope arrives with its own headcount and budget rather than being absorbed.

4. The trust deficit

HQ teams fear the India centre is coming for their jobs, or believe its work is lower quality. The interesting work is retained at HQ, India gets maintenance and tickets, and the centre's quality never gets the chance to disprove the assumption. Self-fulfilling and extremely common.

Mitigation: be explicit at HQ about what India does and does not take; move HQ people into India-facing growth roles rather than displacing them; run exchange visits both directions in the first year; and give India one visible, high-stakes deliverable early — trust shifts on a demonstrated win far faster than on any communication plan.

5. The cost-centre trap

If the centre is measured only on cost per FTE and headcount, it will optimise for those and never build capability. Every budget cycle becomes a cost conversation, comp bands fall behind market, quality degrades, and the centre confirms the prejudice that it is a cheap back office.

Mitigation: change the metrics by month 18 — revenue or product influenced from India, features owned end to end, global roles held, IP generated, time to market. Cost per FTE stays on the dashboard but stops being the headline.

6. Leadership misfire

The wrong GCC head costs 12–18 months and typically the first two cohorts. The three recurring misfires: a delivery manager from an IT services firm who runs the centre as a billable utilisation shop; an operator from a 2,000-person mature site who cannot build from zero; and a strong individual contributor with no experience hiring at scale.

7. Compliance surprises

Contractors misclassified as consultants creating PF and gratuity arrears; no POSH Internal Committee; no crèche at 50+ employees; missed FLA returns; a TP position taken without documentation. None individually fatal; collectively they surface during diligence or an exit and reduce valuation or delay a transaction.

Mitigation: the compliance calendar from day one, outsourced to a specialist firm for the first two years. At $20,000–$60,000 a year it is the cheapest insurance in the build.

8. Real estate over-commitment

A nine-year lease signed in month two for the year-three headcount, at 1:1 seat ratio, with 5–7% annual escalation. If the ramp slows, the client pays for empty floors for seven years.

Mitigation: managed office or short-term flex space until the ramp is proven at 60–70% of plan; then lease, with expansion rights on adjacent space rather than pre-committed area; and a seat ratio derived from the client's own attendance data rather than a published benchmark, since none survives scrutiny.

Exit and wind-down

Every playbook explains how to build a GCC. Almost none explains how to close one, which is exactly why a sceptical CFO asks.

Knowing the exit cost is not pessimism. It is what makes the three-year commitment in section 5 a real commitment rather than a hope, and it changes decisions you take in month one — the lease you sign, the model you enter on, and whether you use an EOR layer at all.

The three exit routes

Routes and timelines per Treelife on closing an Indian subsidiary. The distinction matters commercially: a GCC that traded for three years and holds assets, employees and intercompany balances is not a strike-off candidate. Budget for voluntary liquidation and a year.

The people cost, under the new Labour Codes

This is the largest and most misunderstood line, and it changed on 21 November 2025 when the Industrial Relations Code took effect.

Thresholds and the re-skilling fund per BDO on the Labour Codes effective 21 November 2025 and India Briefing on retrenchment and settlement.

The lease, which is usually the second-largest number

• Lock-in of 3–5 years is typical on a Grade-A office lease, with 12–36 months common depending on tenure and building grade. Indian law sets no limit — it is purely negotiated.
• Notice to vacate is generally 3–6 months in writing.
• Security deposit runs 6–12 months’ rent in stronger markets, and its return is conditional on reinstatement, settled dues and proper notice.
• Exiting inside the lock-in typically forfeits the deposit and triggers rent for the remaining lock-in period.

One point worth knowing, because landlords rarely volunteer it: a lock-in penalty is a penalty clause, and is therefore subject to the reasonableness test under section 74 of the Contract Act. Courts have reduced disproportionate demands — five years of future rent for a two-month early exit, for instance — to the landlord’s actual loss. A punitive exit clause is not necessarily enforceable as written, which is useful leverage in a negotiation that feels one-sided.

The rest of the list

1. Data and IP repatriation. Source code, documentation, customer data and access credentials move back before people leave, not after. Build the inventory while the team is still there and motivated.
2. Transfer pricing in the closure year. A captive is expected to earn its arm’s-length return regardless of the parent’s circumstances. Closure costs sitting in the cost base, and a loss-making final year, both invite challenge.
3. Statutory clean-up before filing. PF and ESI settlement, gratuity payout, final TDS and GST returns, FEMA filings and the FLA return. Arrears discovered mid-liquidation stall the process.
4. Vendor contracts — payroll, facilities, IT, transport — each with its own notice and termination terms.
5. Repatriation of surplus to the parent, which requires the liquidation route and is a principal reason strike-off rarely fits.

The maturity curve

Only about 27% of GCCs reach Portfolio Hub within five years. Nearly three-quarters stall as execution shops.

Zinnov's framework describes four stages — Outpost, Satellite, Portfolio Hub and Transformation Hub. Knowing why centres stall, and what moves them up, is the difference between advising on a GCC and staffing one.

A 60-person centre with real product ownership outranks a 400-person ticket shop.

What actually moves a centre up a stage

Not headcount, and not time. Four things, in order of impact:

1. Transfer of decision rights, not just work. The stage change happens when India decides how, and eventually what, rather than executing a specification. Everything else follows this.
2. A senior technical bench. You cannot own a product with a team whose most experienced person has six years' experience.
3. One India-based person holding a global role. The first time someone in Bengaluru owns a function worldwide, the ceiling visibly breaks and the retention effect on the next tier is immediate.
4. Changing the metrics. A centre measured on cost per FTE will stay a cost centre. Change what the dashboard reports and behaviour follows within two quarters.

Metrics by stage

The mid-market advantage

Mid-market GCCs advance along this curve 1.2x faster than larger centres and show 1.3x higher presence as transformation hubs, despite operating at around 40% of the scale. The mechanism is simple: a $300m-revenue parent cannot afford a 500-person execution-only centre, so India gets real ownership earlier out of necessity. Smaller parents build better centres. A counter-intuitive finding, and one that cuts against the assumption that a GCC is something only large enterprises can run.

The honest tension

As a centre climbs this curve, two things pull against each other. The capability case gets stronger — India owns products, holds global roles, creates IP. The tax case gets weaker, because a centre that genuinely owns IP and bears risk is harder to characterise as a low-risk contract service provider earning a 15.5% cost-plus markup. Few analyses connect these two.

Sources, confidence and maintenance

Not all numbers here are equally solid. Treat them accordingly when quoting them.

Two cautions worth stating plainly. Many cost and salary sources are published by GCC service providers, who have an interest in the numbers looking favourable — the ranges triangulate well, but treat the low end as marketing and the midpoint as realistic. And salary bands move 8–12% a year in India, so any band here has a shelf life of roughly two quarters.

The October 2026 validation run

Seventeen cost and pricing claims were put through a multi-source research pass prioritising independent publishers over GCC vendors, with each extracted claim adversarially verified by three independent votes. Thirteen of twenty-four verified claims were killed. Four of the seventeen cost items survived.

What was independently confirmed

What failed verification

• EY's GCC all-in cost-per-FTE benchmarks — refuted 3-0. The only independent per-FTE anchor available.
• Knight Frank's city-wise average rent table and city-level Grade-A vacancy — refuted 3-0.
• "Employee expenses were 76.2% of GCC cost" and "cost growth is driven almost entirely by salary" — both refuted 3-0.
• The 1.5 headcount-per-seat hybrid benchmark — refuted 3-0.
• Four separate salary-increment projections, including two Aon-attributed figures from a superseded October 2025 vintage. Use the February 2026 numbers; do not quote the 9.0% / 9.5% pair.

What has no evidence either way

Setup cost by team size, year-one cost per FTE, steady-state cost per engineer, the 40–60% US savings claim, managed-office per-seat pricing, salary bands by role, statutory loading totals, EOR pricing and the entity crossover point, entity standing overhead, lease escalation norms, productivity ramp curves, recruitment fee norms, and the cash value of state incentives. Thirteen items, no surviving independent source. Their absence from the evidence base is itself a finding — do not quietly refill them from vendor material.

What to re-validate, and when

The research agenda

Five open questions, each one nobody in the market has answered:

1. Within-city basic-versus-premium fit-out spread, and the real seat-to-headcount ratio under hybrid. The highest-leverage assumption in any rent or fit-out calculation, and the published benchmark failed verification.
2. An independently derived all-in cost per FTE. The largest gap in the public evidence base.
3. Grade-A rents and vacancy for genuine tier-2 candidates — Coimbatore, Kochi, Indore, Jaipur, Chandigarh, Mangalore. No priced source covers them.
4. Documented evidence of state GCC incentives actually disbursed, not merely announced.
5. The 2027 India projection alongside US increment rates, without which the arbitrage erosion rate cannot be computed.