The Future of GCCs: From Cost Centers to Value Centers
The Global Capability Center is being rebuilt around a different question. Not how cheaply can this function run offshore, but what can this team originate that we could not build anywhere else.
The short answer. The future of GCCs is smaller, faster to launch, and measured on value created rather than cost saved. The 500 person campus built over twelve months is being replaced by 10 to 200 person, mandate-specific centers that reach full velocity in a single quarter. AI is the accelerant: it compresses the headcount a given mandate requires, and it has become the single fastest growing GCC mandate in its own right.
The Global Capability Center is not a new idea. India has hosted them for more than twenty five years, and the largest are among the most sophisticated technology operations anywhere in the world. What is changing is not the destination or the talent. It is the entry price.
For two decades, a GCC was something only large enterprises could credibly attempt, because the fixed overhead of a legal entity, a compliance function, a facilities operation and an HR apparatus only made sense spread across hundreds of people. That constraint is what produced the campus. Remove the constraint, and the campus stops being the natural shape of the thing.
This guide covers the three shifts driving that change, what each generation of the GCC model optimised for, and what a company evaluating the decision in 2026 should actually plan around.
The scale of the shift
India’s GCC sector is growing in count faster than it is growing in average size. That distinction is the whole story.
What is actually changing
These are not independent trends. Each one makes the next more viable.
Speed became the binding constraint
A nine to twelve month build cycle is no longer competitive when a direct competitor can stand up an equivalent capability in a quarter. The company that waits is not being careful. It is conceding a two to three quarter head start on the same roadmap.
AI compressed the headcount maths
Agentic tooling and modern engineering practice mean a fifteen person pod can credibly own work that once needed sixty people. That is the arithmetic that makes a small center viable where it previously was not.
Boards started asking a different question
Cost savings alone makes a team permanently re-biddable. Leadership teams increasingly ask what the center has built, owned and de-risked, which is a fundamentally different standard to hold a team to.
GCC 1.0, 2.0 and 3.0 compared
Each generation solved the problem the previous one created.
| Dimension | GCC 3.0 (now) | GCC 2.0 (2010s) | GCC 1.0 (2000s) |
|---|---|---|---|
| Primary mandate | Specific capability: product, AI, innovation | IT and shared services at scale | Back office and transaction processing |
| Typical size | 10 to 200 people | 1,000 to 5,000 people | 500 to 3,000 people |
| Time to launch | 3 to 4 months | 9 to 15 months | 12 to 24 months |
| Measured on | Value created and IP owned | Throughput and SLA compliance | Cost per transaction |
| Talent profile | Senior engineers, AI specialists | Engineers and analysts | Process operators |
| Who can afford one | Growth stage upward | Large enterprise | Large enterprise |
| Entity and compliance | Carried by a partner | Carried in house | Carried in house |
For the full breakdown of the current generation see What Is a Nano GCC, and GCC 3.0: From Captive Centers to Nano GCCs.
Why AI changed the size of a viable team
The traditional GCC needed scale for a structural reason: fixed overhead. A captive legal entity, a facilities team and a dedicated compliance function all have a floor cost, and that floor has to be amortised across enough headcount to disappear into the unit economics.
Agentic AI attacks the problem from the other side. If a small, senior team supported by good tooling can own the throughput that previously required a large one, then the headcount needed to justify the overhead never has to be reached in the first place. Combine that with a partner already carrying the entity, and the floor cost problem is solved twice over.
AI is also, separately, the fastest growing mandate. A large share of new centers are not being built to run existing processes more cheaply. They are being built because the parent company cannot hire senior AI engineers at home quickly enough.
- Smaller teams can credibly own larger mandates
- Partner-carried overhead removes the scale floor entirely
- AI capability is now a launch mandate, not a later phase
- Senior talent availability, not cost, is often the deciding factor
How the next 24 months are likely to unfold
A realistic planning view rather than a forecast. Each stage is already visible in the market today.
Mandate-first centers become the default
New centers are increasingly scoped around one named capability rather than a headcount target. The question moves from how many people to which problem, and team size follows from the answer.
AI mandates overtake support mandates
A growing share of new center launches are agentic AI and product engineering teams rather than process or support functions, driven by domestic hiring constraints rather than by cost.
Value measurement becomes standard practice
Boards increasingly ask for outcome reporting: IP owned, time to market improvement, roadmap risk retired. See the Value Generation Framework for the specific measures.
Cost transparency becomes an expectation
All-in cost modelling replaces rate card comparison in serious procurement processes, because the gap between the two is where most failed business cases originate. See True-Up Cost Methodology.
The practical implication
If you are evaluating a capability center today, the useful conclusion is not that the traditional model is obsolete. For a genuinely large, steady state function it remains the right answer.
The conclusion is that the model is no longer the only option, and that the default has moved. A company needing one specific capability live this quarter now has a route that did not exist a few years ago, and choosing the older shape by default carries a real cost in time.
- Scope the mandate before you scope the headcount
- Assume a quarter, not a year, as the planning horizon
- Set value measures at launch, not retroactively
- Model all-in cost before comparing any two options
The mistake we nearly made was assuming we needed to be a certain size before an India center made sense. That was true when we last looked at it. It stopped being true, and nobody sent us a memo.
Common questions about where GCCs are heading
What is GCC 3.0?
GCC 3.0 is the current generation of Global Capability Centers: smaller, mandate-specific teams such as Nano and Micro GCCs that launch in a quarter and are measured on value created, rather than the large-scale, cost-arbitrage-focused captive centers of earlier generations.
Are traditional large GCCs going away?
No. For genuinely large, steady state functions a large captive center remains appropriate. What has changed is that scale is no longer a prerequisite for having a center at all, so the large format is now one option among several rather than the only route.
How is AI changing GCC strategy?
In two ways. It reduces the headcount a given mandate requires, which makes small centers viable, and it has become one of the most common mandates in its own right as companies struggle to hire senior AI engineers domestically.
Is cost still the main reason companies build GCCs?
Increasingly not. Talent availability and speed to capability are now frequently the primary drivers, with cost structure a supporting rather than deciding factor. This is a meaningful change from the previous decade.
What size should a new center be in 2026?
Most new centers launch at 10 to 40 people and grow toward 10 to 200 as the mandate proves out. Sizing to the mandate rather than to a target is the current best practice.
How long should a launch realistically take?
Three to four months to full operational velocity is achievable when compliance, facilities and hiring run in parallel. Nine to twelve months reflects a sequential build with the entity carried in house.
Does India remain the leading destination?
Yes, on talent depth, ecosystem maturity and cost. See Why India for the detailed case, including how the timezone overlap is typically used.
What is the biggest risk in this shift?
Underestimating compliance. A smaller center does not mean smaller legal obligations, and entity or labour law mistakes are expensive to unwind regardless of team size.
Continue reading
Where to go deeper on each of the shifts covered above.
Where the new mandates are concentrating
Six areas account for the large majority of centers launched in the current cycle. Each is a capability a company struggled to build at home.
Agentic AI and applied ML
Teams building autonomous agents, evaluation harnesses and production ML pipelines. Structured around a specific architecture lead plus pipeline engineers rather than a pool of generalists, which is what separates shipped capability from demos.
Product engineering pods
Cross functional squads of four to eight engineers with a tech lead, embedded directly in the parent company sprint process and owning a defined roadmap area outright rather than taking tickets.
Proof of concept and validation
Small time-boxed teams that take a concept from specification to working prototype, ending in an explicit decision point: scale in place, hand off to the core team, or stop cleanly.
Innovation and IP origination
Teams trusted to identify problems as well as solve them, running their own discovery and bringing finished capability back to the business. The most advanced form of the model.
Platform and infrastructure
Cloud, developer experience and reliability engineering teams that free the parent company core team from undifferentiated heavy lifting without offshoring product decisions.
Security and compliance engineering
Dedicated capacity for the security, privacy and regulatory engineering work that is chronically under-resourced when it competes with feature delivery for the same engineers.
Months from decision to a team at full velocity
Ranges reflect typical engagements. The compression comes from running compliance, facilities and hiring in parallel rather than sequentially.
Why the maths changed
Fixed overhead stopped requiring scale
The legal entity, compliance function and facilities operation used to have to be paid for by headcount. When a partner already runs them across many clients, the floor cost disappears from your unit economics entirely.
Senior talent became the scarce input
For AI and platform roles the binding constraint is availability, not budget. A market that can supply four senior engineers this quarter is worth more than one that can supply them eventually at a slightly lower rate.
Time to capability entered the business case
A capability live two quarters earlier compounds. That value rarely appeared in older GCC business cases because the timeline was assumed rather than treated as a variable you could influence.
Cost comparison got more honest
Serious business cases now model benefits, attrition backfill, facilities and compliance on both sides rather than comparing a headline offshore rate against a fully loaded domestic salary.
Three patterns defining the current cycle
Representative deployment patterns drawn from how the model is typically applied, not named client accounts.
The deferred roadmap area
A company with a fully committed domestic team stands up a dedicated pod to own a roadmap area that had been deferred twice, without pulling focus from existing commitments.
The capability that could not be hired
Unable to hire senior AI engineers domestically within the roadmap window, a company builds a focused agentic team with an architecture lead, pipeline engineers, MLOps and an embedded product owner.
The bet worth de-risking first
Rather than committing core team bandwidth to an unproven idea, a small time-boxed team validates it to working prototype and returns a clear scale, hand off or stop decision.
Planning your capability center for what comes next?
Tell us the mandate you need owned. We will come back with a team shape, a realistic timeline and an all in cost model you can take to your board.