Given the constraints of capital, computing capacity, energy and infrastructure, pursuing “scale in artificial intelligence for its own sake” is neither efficient nor necessary, and India should aim for a bottom-up, multiple sector-specific approach to ensure it pays dividends, the Economic Survey 2025-26 has said.
The survey flagged high concentration of frontier large language model development in the hands of a few major firms, and export controls imposed on the most-advanced processors required for scaling up frontier model development, as major entry barriers. The most widely deployed AI models are also proprietary and opaque, limiting transparency around their training data, internal logic and update mechanisms, it said.
The ensuing fundamental asymmetry may see most countries end up participating in AI primarily as users, while a few shape the technology's trajectory. "Attempting to close this gap would involve prohibitive fiscal costs towards what is increasingly becoming an unsustainable approach to AI development," the survey said.
As a result, India would have to choose between expending scarce resources to chase frontier-scale models and deploying those resources more effectively towards domain-specific AI systems aligned with domestic economic priorities, it said.
On the other hand, it called data centres a double-edged sword which now guzzles even more power and water after the advent of AI data centres, competing directly with demand from households and industry. But it recognised that India would need to beat emerging hubs such as Malaysia, Japan and Vietnam to set up more data centres given the country currently hosts only 3% of the 11,000 data centres worldwide, even as it generates nearly 20% of global data. Data centres will emerge as geostrategic leverage akin to critical minerals, it said. Future success in manufacturing will depend on India's ability to embed itself into global value chains as a high-tech, highly productive manufacturing hub, it said. "Medium and high-technology industries expanded by 1.7% and 1.4% in Q2 and Q3 of 2025, respectively, significantly outpacing stagnant low-tech While noting job losses due to AI may not arrive overnight, the survey argued that a lag in AI adoption risks was hollowing out the sector's “core value proposition”.
The survey flagged high concentration of frontier large language model development in the hands of a few major firms, and export controls imposed on the most-advanced processors required for scaling up frontier model development, as major entry barriers. The most widely deployed AI models are also proprietary and opaque, limiting transparency around their training data, internal logic and update mechanisms, it said.
The ensuing fundamental asymmetry may see most countries end up participating in AI primarily as users, while a few shape the technology's trajectory. "Attempting to close this gap would involve prohibitive fiscal costs towards what is increasingly becoming an unsustainable approach to AI development," the survey said.
As a result, India would have to choose between expending scarce resources to chase frontier-scale models and deploying those resources more effectively towards domain-specific AI systems aligned with domestic economic priorities, it said.
On the other hand, it called data centres a double-edged sword which now guzzles even more power and water after the advent of AI data centres, competing directly with demand from households and industry. But it recognised that India would need to beat emerging hubs such as Malaysia, Japan and Vietnam to set up more data centres given the country currently hosts only 3% of the 11,000 data centres worldwide, even as it generates nearly 20% of global data. Data centres will emerge as geostrategic leverage akin to critical minerals, it said. Future success in manufacturing will depend on India's ability to embed itself into global value chains as a high-tech, highly productive manufacturing hub, it said. "Medium and high-technology industries expanded by 1.7% and 1.4% in Q2 and Q3 of 2025, respectively, significantly outpacing stagnant low-tech While noting job losses due to AI may not arrive overnight, the survey argued that a lag in AI adoption risks was hollowing out the sector's “core value proposition”.
