What Is a Global Capability Center? A 2026 Guide for CTOs

What is a global capability center?

A global capability center is an offshore office that a company owns and staffs directly, running engineering, data or operations work for the parent business. The people are employees, not a vendor's headcount, and the roadmap stays in-house.

The older name was the captive center, and it described the same arrangement: a wholly owned subsidiary in a lower-cost country, set up because the work was continuous rather than project-shaped.

What changed is the mandate. Early captives handled maintenance, support and regression testing. Current ones own products end to end, including the parts the parent company cannot hire for at home.

Roughly 80% of centers launched in 2026 name AI and machine learning as their core mandate, according to the NASSCOM-Zinnov India GCC report for 2026. That is a different hiring problem from staffing a support desk, and it is the main reason the model is back in front of boards that rejected it a decade ago.

What does GCC mean in business terms?

GCC stands for global capability center. In practice the term means an incorporated offshore entity, owned by the parent, whose staff sit on the parent's payroll and whose output is product rather than a deliverable against a statement of work.

The acronym gets stretched. Plenty of vendors now sell "GCC-as-a-service" for arrangements where the vendor still employs everyone and still books the margin.

Three tests settle it. Who signs the employment contracts. Whose P&L carries the cost. Who decides a senior engineer's promotion.

If the answer to any of those is the vendor, the arrangement is outsourcing with better branding. That is not a criticism of outsourcing, which is often the correct choice. It is a warning about buying one thing while believing you bought another.

How big is India's GCC market in 2026?

India hosts 2,117 GCCs generating USD 98.4 billion and employing about 2.4 million professionals in FY2026, according to NASSCOM and Zinnov. It is the largest concentration of owned offshore capability anywhere.

The projection attached to that research puts India above 2,500 centers by 2030, employing 2.8 to 2.9 million people.

Two things are worth pulling out of those numbers before they become background noise. The first is that 2.4 million people is larger than the entire technology workforce of most countries, which means the supply side is no longer the constraint it was in 2010.

The second is that USD 98.4 billion divided across 2,117 centers gives an average well under USD 50 million in annual output per center. The median GCC is not a thousand-person campus. It is a department.

Who is actually opening these centers?

The mix is broader than the headlines suggest: 506 Forbes Global 2000 companies run a GCC in India, alongside 583 mid-market and 504 PE-backed centers, per NASSCOM-Zinnov.

Read that ordering again. Mid-market centers outnumber Global 2000 centers. The model is no longer restricted to companies with a corporate development team and a tax structuring practice on retainer.

PE-backed centers are the more interesting signal. A sponsor with a five-year hold period does not fund an entity setup for sentiment. They fund it because owned engineering capacity changes the multiple at exit in a way a vendor contract does not.

That is the honest commercial argument for the model, and it is a balance sheet argument rather than a cost argument.

GCC vs outsourcing: what actually changes?

Outsourcing buys delivery capacity from a vendor who owns the team. A GCC buys nothing; it builds the team. The difference shows up in three places: who employs the people, who holds the IP, and who sets the roadmap.

Everything else that gets argued about, rates, time zones, communication overhead, is true of both models and is not a reason to choose between them.

Who owns the people, the IP and the roadmap?

People first. In a vendor arrangement, your best engineer's career ladder belongs to the vendor, and so does the decision to move them to another account. In a GCC, you decide.

IP second. Vendor contracts assign IP on delivery, which works until you need the reasoning behind a decision made eighteen months ago by someone who has since left the vendor.

Roadmap third, and this is the one that decides most cases. Vendor teams optimise for the scope you wrote down. Owned teams can tell you the scope was wrong.

That last property is what companies are really buying. It is also the hardest to get, because it requires local leadership senior enough to disagree with head office.

How do the four delivery models compare?

Four arrangements sit on a spectrum between pure outsourcing and a wholly owned center. The definitions follow Morgan Lewis in its sourcing practice: an ODC stays vendor-owned, a captive is client-owned, and Build-Operate-Transfer bridges the two.

The column that gets ignored during vendor selection is the last one, exit cost. It is what you are actually paying for when you choose project outsourcing, an ODC or a BOT over a GCC: a cheaper exit today in return for a harder one to walk away from later.

What does the global capability center model look like in practice?

Three routes exist. Build the entity directly and hire from scratch. Run Build-Operate-Transfer, where a partner builds and operates the center then transfers team, tooling, process and entity to you. Or start with an ODC and convert it later.

Direct build gives the cleanest ownership and the slowest start. You are incorporating, registering for payroll, appointing directors and hiring a site leader before a single line of code exists.

BOT compresses the first eighteen months by renting someone else's operating capability, then paying a transfer fee to take it over. The failure mode is a transfer that moves the entity but not the institutional knowledge, because the people who held it were the partner's staff and chose not to move.

The ODC-first route is the cheapest way to find out whether the offshore work is actually continuous, which is the question that should decide the whole thing.

What does a capability center cost to run?

Salary arbitrage is the smallest line in year one. Entity incorporation, transfer pricing compliance, payroll infrastructure, a site leader who can hire, and a lease all land before productive output does.

Those costs are close to fixed. They are the same whether the center holds fifteen engineers or ninety, which is why the model rewards scale and punishes hesitation.

The variable saving is real but narrower than the pitch decks claim. EY puts the gap at a 10% to 35% lower cost of living than the nearest tier-1 location, and reports that attrition rates in tier-2 cities "have been observed to be up to 10% lower than Tier-1 locations". EY's own list of Indian tier-2 cities with GCC operations names Chandigarh.

Two caveats worth repeating to your CFO. That is a cost-of-living gap rather than a straight salary cut, and it narrows as a city fills up. Build the business case on retention and ownership, not on a wage difference that is closing while you are still hiring.

Why is a global capability centre in India the default choice?

Depth of supply, and nothing more romantic than that. 2.4 million people already do this work, which means a hiring manager is recruiting from an existing pool rather than creating one.

The second reason is management maturity. Twenty years of captives means engineering directors who have already run a distributed team through a reorganisation, an audit and a platform migration.

The third is that the AI mandate has to go where the people are. You cannot staff a machine learning platform team out of a city with four candidates in it, regardless of the cost model.

Countries competing on cost alone have not solved the second problem, and it is the expensive one.

When is a global capability center the right answer?

Five conditions, and you want most: continuous work, a multi-year roadmap, enough offshore headcount to amortise fixed costs, IP you'd rather not assign by contract, and a leadership bench with real decision rights in another time zone.

Our rule of thumb after nineteen years of running offshore delivery: below roughly 40 offshore engineers on work that will still exist in three years, the fixed costs do not pay back.

Above that, the arithmetic flips quickly, and the ownership benefits compound rather than repeat.

The condition most often missed is the last one. A center that has to ask head office before changing a schema is an expensive way to buy the same thing an ODC provides.

When is a GCC the wrong answer?

A 20-person product team does not need one. Neither does a company before product-market fit, a team whose roadmap changes every quarter, or a business whose offshore need is one migration with a defined end date.

The cost of being wrong is asymmetric. Ending a vendor contract takes a notice period. Unwinding an entity takes redundancy payments, statutory filings and a year.

There is also a quieter failure. A center built for a mandate that evaporates does not close, it drifts, and eighteen months later the parent is paying for a team maintaining work nobody sponsors.

Continuity is the real test. If you cannot name the work this team will be doing in three years, you are not ready to own them.

What to build instead below that threshold

An extended engineering team, a fractional leadership arrangement, or a funded first build. All three give you offshore capacity without an entity, and all three are reversible.

If the question is whether an idea is worth a team at all, Moonship scopes and builds a defined MVP in 30 days, starting at $2,999 for a fixed scope, which settles the question faster than a location study will.

If the gap is engineering leadership rather than hands, CTO-as-a-Service, billed hourly from $100, buys specific architecture decisions, roadmap reviews or a technical due-diligence read, without committing to a full-time hire.

If the work is real and continuing but still under that 40-engineer line, a dedicated product development team is the same operating shape as a small center, minus the compliance calendar.

Why is Mohali on the GCC map now?

Because the centers stopped going only to metros. Nearly a quarter of India's new GCC units in the past year landed in emerging cities, Mohali among them, according to the same NASSCOM-Zinnov research.

The talent argument is the one that holds up: the Chandigarh tricity produces over 40,000 fresh graduates a year, according to Invest Punjab, and already has a working base of technology professionals who did not have to relocate to get experience.

The infrastructure exists rather than being promised: Rajiv Gandhi IT Park in Chandigarh, Quark City in Mohali, and Panchkula IT Park.

The Government of Punjab partnered with Zinnov as State Partner at Confluence 2026, presenting through Invest Punjab in a session titled "The Location Playbook: Mohali as North India's Emerging GCC Hub", reported by Business Standard in August 2026.

What is Punjab offering?

Money, mostly, and the structure is unusually specific. Punjab's Industrial and Business Development Policy 2026 covers IT, ITeS, data centres and GCCs, targeting roughly Rs 75,000 crore of investment, published by Invest Punjab.

The line that matters to a finance director is the employment subsidy of Rs 7,500 per employee per month, alongside rental and capital subsidy.

Treat that as a discount on the variable cost, not as the reason to choose a city. Subsidies have expiry dates and attrition does not.

The honest counterweight: senior and niche hiring is harder outside Bengaluru, Hyderabad and Pune, and a first site leader may need relocating. Budget for that rather than discovering it in month four.

How does Viithiisys fit into this?

Viithiisys is an engineering studio in Mohali that has run a dual-shore model since 2007: delivery from Mohali, client-facing presence in Markham, Ontario.

That is the same shape a small capability center takes, without the entity, in the city Punjab is now putting on the GCC map.

The specifics: founded 2007, nineteen years in, 20+ engineers, designers and strategists, 500+ projects shipped across six countries, with client work including Paytm, Snapdeal, IKEA, Nestlé, Shiprocket and Vikram Solar.

Recent builds are on the case studies page: skin analysis and an AI assistant wired into Conscious Chemist's website and CRM, client-confirmed at +15%; a voice-first coaching layer inside Fitelo's existing app and coach workflow; driver communication and fleet operations on one platform for Milo; and Vizitor taken from a seed-stage idea to a security-first workplace product.

Where should you start?

Not with a location study. Start by establishing whether the offshore work is genuinely continuous, because that single answer decides between a vendor contract and an entity.

A practical first step is to map where work currently stalls, who owns each handoff, and which parts of the roadmap keep slipping for reasons nobody can name. Our broken workflow assessment does exactly that, and it costs nothing to find out that the answer is "not yet".

If the answer turns out to be an extended team rather than a center, that is a cheaper and more reversible outcome. Tell us what you are building and we will say plainly which one your situation calls for.