The history of Global Capability Centers: from cost center to capability engine

The Global Capability Center reads like a recent invention. Offshore engineering teams, embedded analysts, a Fortune 500 closing its books from Bengaluru and Manila. The model is actually forty years old. Its history has a clear shape, and the shape tells you where it goes next.

We have already covered what a Global Capability Center is and why it stayed an enterprise tool for so long. This is the other half of the story: where the model came from, how it evolved era by era, and the one direction it has moved since 1985.

1985: a satellite link between Bengaluru and Dallas

Industry histories date the first Global Capability Center to 1985, when Texas Instruments opened a wholly owned design center in Bengaluru. TI was the first multinational to run an R&D operation in India, and the first company in the country licensed to operate its own dedicated satellite link. The link connected a small team of Indian engineers directly to headquarters in Dallas, and their work fed chip-design tools used across TI's global product line.

The same year, with less fanfare, Citibank stood up Citicorp Overseas Software at SEEPZ in Mumbai, a wholly owned subsidiary writing the banking software that ran the bank's worldwide operations. Industry chroniclers count it as India's first captive center. Both companies went for the same reason. India's engineering institutes had been producing world-class graduates for decades, and almost no global company had found a way to employ them where they lived.

Notice what it took to participate. Government approval. Your own satellite link. Years of patience before the output justified the plumbing. In 1985 the model was available to a company with the balance sheet of Texas Instruments and to almost nobody else.

The captive era and the Y2K wave (1991 to 2005)

The model spread slowly for its first decade, then all at once, and it spread as a cost story. India liberalized its economy in 1991 and the regulatory friction that had made the pioneers exceptional fell away. Then the Y2K remediation effort of the late 1990s sent a generation of Western companies offshore for the first time and proved that distributed technical work could run at scale. Indian software exports crossed $1 billion in 1997 and reached $6.2 billion within four years.

The captives multiplied on the back of that proof. GE built GECIS in Gurgaon in 1997 to run GE Capital's back office, processing car loans and credit-card transactions. Goldman Sachs arrived in Bengaluru in 2004 with roughly 300 people doing IT and support work. The industry's name for these operations, the captive center, was honest about the era. Headquarters specified the work. The center executed it. The business case was a labor-rate table, and for a while almost nobody measured anything else.

The entry price was still enterprise-sized. A foreign entity, a building, a management layer, a multi-year ramp. What changed is that the model stopped being an experiment. There was a playbook now.

The crisis of faith: why companies sold their captives (2004 to 2009)

The companies that sold their captives in the mid-2000s sold them because they had built them as cost lines, and a cost line gets managed like one. GE sold a majority stake in GECIS to two private equity firms in 2004. The operation was renamed Genpact and began selling its services to other companies. Citigroup sold its India business-process captive to Tata Consultancy Services in 2008 for roughly half a billion dollars, then sold its India technology arm to Wipro the same year, in the teeth of the financial crisis.

The divestment wave looked like a verdict on the model. It reads differently with distance. The companies that had built pure execution capacity sold it, because execution capacity is a commodity and commodities are for sale. The companies that had pushed real capability into their centers held on. Nobody sold a center they could not function without.

The capability turn (the 2010s)

The centers that came through that sorting kept moving toward the core of the business, and by the mid-2010s the work itself had changed. The Goldman Sachs office that opened with 300 support staff is today the bank's largest office outside New York. By 2023 Bloomberg was reporting roughly 8,000 people there, and they are quants and engineers building the systems the bank trades and manages risk on. Target's Bengaluru center, opened in 2005 as an IT support team, now works on merchandising analytics, pricing, and supply-chain systems for the parent company.

Somewhere in this stretch the industry retired the word captive and started saying Global Capability Center. Renames are usually marketing. This one was reporting. The thing inside the buildings had changed, and the name caught up with it.

The model moves down-market (the 2020s)

The model is growing faster today than at any point in its history, and it is growing because the price of entry keeps falling. In its 2024 landscape report with Zinnov, NASSCOM, the Indian tech-industry association, counts more than 1,700 GCCs in India employing over 1.9 million professionals, a sector that generated $64.6 billion that year and is projected to reach roughly $100 billion by 2030.

The newest entrants are the real story. A 2025 NASSCOM and Zinnov analysis counts more than 480 mid-market centers in India. Over 45 of them arrived in the last two years alone, nearly 35 percent of all new GCC additions in that stretch. Cloud infrastructure, modern collaboration tools, and a mature ecosystem of local enablers have stripped out most of the fixed costs that once made the model an enterprise privilege. Companies a fraction of GE's size are now running a play that GE needed a decade and thousands of seats to prove.

The model was born enterprise-sized. It has spent forty years working its way down-market. The middle market is the last stop on the line.

What history says happens next

Line up the eras and the direction is hard to miss. In 1985 the model required government approval and your own satellite link. In 2005 it required a foreign entity, a building, and a thousand seats to make the overhead pencil. Today it fits a mid-market enterprise running a few hundred seats. Every era made the model smaller, faster to stand up, and closer to the core of the business.

The line does not stop at a few hundred seats. The last market it reaches is the true middle market: PE-backed companies between $50 million and $500 million in revenue, where the right team is ten seats rather than ten thousand, and where even today's shrunken version of the overhead never pays for itself. Most of those companies have only ever experienced offshore through the vendor model, and the disappointment that follows has its own anatomy.

History already shows what this next step looks like, because it is the same step every era took. Keep the architecture. Shed more of the wrapper. The rebuilt version is the Mini-GCC: the same embedded structure the enterprises spent forty years proving, without the entity, the real estate, or the multi-year build. Kayana creates Mini-GCCs for PE-backed middle-market companies. That is the chapter of this history we were built for.

Quick answers

The category argument continues in the Mini-GCC model.