Global Capability Centres are the in house offshore arms of multinationals. India is now the dominant host as the work inside them shifts from back office operations to AI strategy.
India hosts roughly half the world's Global Capability Centres. The number is striking, but the more consequential shift is what those centres now do.
A Global Capability Centre is the in-house offshore arm of a multinational, owned by the parent company, not an outside vendor. Two decades ago these were largely back-office operations: processing transactions, handling customer calls, entering data. Today the same buildings house AI engineering teams, machine learning platform groups, and product managers who set priorities for global AI rollouts.
India's Chief Economic Adviser, V. Anantha Nageswaran, made the half-the-world claim in remarks reported across Indian business press on 9 July 2026, framing the milestone as a structural economic shift rather than a hiring statistic (Moneycontrol). The figure itself is a CEA assertion drawn from his characterisation of the sector; independent base counts come from industry trackers such as Nasscom and Zinnov, whose joint studies have tracked the GCC population for years.
The more revealing datapoint sits in a report Nasscom and Zinnov published days earlier. Indian GCCs have moved from executing AI projects handed down by headquarters to leading the AI mandate themselves: owning the roadmap, choosing the models, deciding the deployment cadence (Business Standard). For a US or European multinational, that means the question of where its AI strategy is conceived has moved thousands of miles from headquarters.
Inside a contemporary GCC, the work looks different from the call-floor image most readers carry. Teams build and fine-tune large language models, run retrieval pipelines over proprietary corpora, maintain model evaluation harnesses, and integrate AI features into the parent's product surfaces. Some centres now run inference at scale for the parent's global user base, not just internal tooling. A handful of the largest GCCs publish research papers and hold seats on the parent's AI governance committees. The relocation is not just labour arbitrage. It is decision-making relocation.
Nageswaran used the same appearance to flag the pressure that comes with that shift. AI, he said, "elevates the value of professionals in India rather than replacing them," but warned there was "no time for complacency" (ANI, The Hindu). The worry is straightforward. If India's edge in GCCs rests on cost arbitrage plus a deep engineering pool, automation of the lower end of that pool erodes the cost case, while the higher end competes directly with Bay Area, London, and Tel Aviv salaries. A CNBC TV18 read of the same speech framed it as the AI exposure of a low-skill model (CNBC TV18).
The second pressure is structural. Once a GCC sets the AI roadmap for a global enterprise, it inherits the regulatory and reputational exposure of that role. An AI deployment that mishandles EU customer data, fails an internal audit, or breaks a model risk control becomes a problem owned in Bengaluru, not Boston. The government's information arm echoed the same framing in its own release (PIB).
Nageswaran's call to industry was concrete. Government has done its part through budget allocations and policy support; the next move belongs to companies that must invest in skilling and innovation to defend the position (Economic Times, Rediff). The framing implies a quiet handoff: from state-supported expansion to private-sector capability investment.
The watch items are dated. The Nasscom-Zinnov study will refresh its GCC census in 2027. Several of the largest GCC operators are renegotiating their AI governance arrangements with headquarters as inference and model ownership shift inward. If the next census shows the share of GCCs that own their parent's AI roadmap crossing a clear majority, the "back office to AI hub" framing stops being metaphor and becomes bookkeeping.