India should resist the temptation to compete in the global race to build the most powerful artificial intelligence models and instead focus on developing smaller, sector-specific AI tools aligned with its economic priorities, the Economic Survey 2025-26 has recommended.
In a clear shift away from the AI arms race that has led companies such as Google, Microsoft, OpenAI and Apple to invest billions into building large capacities for foundational models, the survey argues that India’s strengths and constraints demand a different strategy — one that focuses on practical use cases rather than headline-grabbing frontier models.
“India’s access to cutting-edge computer infrastructure is limited, financial resources for large-scale model training are scarce, and private participation in foundational AI research remains relatively muted compared to global leaders. These constraints render the pursuit of foundational models as the centrepiece of an AI strategy challenging,” the survey says.
The alternative approach proposed focuses on application-specific, small AI models designed to solve concrete problems in key sectors. Such models are “significantly more computationally efficient, easier to fine-tune, and capable of running on locally available hardware, such as smartphones or personal computers, making them better suited to India’s existing infrastructure base,” the survey says.
While experts welcomed the survey’s focus on application-specific models, Rohit Kumar, founding partner at public policy firm The Quantum Hub, cautioned against relying solely on downstream, task-specific tools. “Frontier models play a critical role in advancing capabilities, reducing long-term dependence, and enabling spillovers across sectors. AI is a horizontal, general-purpose technology—much like the internet—and cutting-edge innovation will ultimately shape market dynamics. To remain globally competitive, India should not opt out of frontier investments altogether; sequencing them may be a better approach,” Kumar said.
{{/usCountry}}While experts welcomed the survey’s focus on application-specific models, Rohit Kumar, founding partner at public policy firm The Quantum Hub, cautioned against relying solely on downstream, task-specific tools. “Frontier models play a critical role in advancing capabilities, reducing long-term dependence, and enabling spillovers across sectors. AI is a horizontal, general-purpose technology—much like the internet—and cutting-edge innovation will ultimately shape market dynamics. To remain globally competitive, India should not opt out of frontier investments altogether; sequencing them may be a better approach,” Kumar said.
{{/usCountry}}Ankit Bose, head of AI at Nasscom, said the industry is already pivoting towards the Survey’s recommended path. “Indian IT firms are increasingly investing in right-sized, domain and task-optimised models in sectors like BFSI, healthcare and manufacturing, where accuracy and cost-efficiency matter more than raw model size,” Bose said. He added that while there is pressure to “go big” due to global narratives, scale alone does not translate into value for most enterprise use cases.
The survey recognises India as a late entrant into the AI race but frames this as an opportunity. “Being a late mover gives India the benefit of hindsight, allowing policy and innovation choices to be shaped with greater intentionality,” it says. The survey also warns that control over critical building blocks is increasingly in the hands of a few players. This concentration could threaten India’s long-standing IT services advantage, it says. “It also raises a substantial question about the future of India’s IT sector, as firms that once relied on India’s comparative advantage to handle a bulk of their work may no longer need to do so. It risks hollowing out India’s core value proposition if adaptation lags.”
To stay competitive, incremental change will not suffice, the survey argues. “If the country is to sustain its competitive edge in IT, a comprehensive evolution is necessary, one that takes full advantage of the potential embedded in AI development and deployment.”