Key Points:
- India backs Small Language Models to cut costs and protect data.
- SLMs run on smartphones, avoiding heavy data centers.
- Focus is on local solutions and job-friendly AI.
India will prioritize Small Language Models over large AI systems to reduce costs, manage energy use, and protect data sovereignty, the India’s Economic Survey 2026 said Wednesday, outlining a bottom-up strategy tailored to local needs and constraints.
India is betting on smaller, task-specific artificial intelligence models rather than competing head-on with the massive Large Language Models dominating the West, according to India’s Economic Survey 2026 tabled in Parliament on Wednesday.
The survey said the government has adopted a “bottom-up approach” to AI that favors Small Language Models, or SLMs, which are cheaper to run, easier to fine-tune, and capable of operating on local devices such as smartphones. The strategy aligns with the Atmanirbhar Bharat, or self-reliant India, initiative.
“By focusing on application-specific, efficient models, India aims to move from being a passive consumer of AI to a source of global reliability and value,” the survey said.
Survey Cites Resource Limits in Global Race for Large AI Models
The document highlighted structural constraints that make replicating Western-style Large Language Models difficult for India, including shortages of high-end computing chips, energy demands, and water use.
Training and deploying advanced LLMs is increasingly capital- and compute-intensive, the India’s Economic Survey said, with access to high-performance graphics processing units limited by export controls and concentrated global supply chains.
It warned that large AI data centers can consume energy comparable to a rocket launch each day and require up to 2 million liters of water daily, putting pressure on infrastructure already under strain.
“Pursuing scale for its own sake carries high opportunity costs and risks financial contagion from highly leveraged infrastructure bets,” the survey said.
In contrast, SLMs can perform focused tasks efficiently and reduce dependence on always-on data connectivity. Modern smartphones, the report noted, have sufficient processing power to run such models locally, allowing AI adoption without a matching expansion in resource-heavy data centers.
Bottom-Up AI Strategy Targets Inclusion, Local Problems
India’s approach leverages its large pool of engineers and one of the world’s fastest-growing open-source developer communities, the survey said. Open models are steadily narrowing the performance gap with proprietary systems, lowering barriers for domestic innovators.
Rather than relying on general-purpose AI, the government is promoting problem-driven applications designed around local conditions and public needs.
Examples cited include linguistic inclusion through platforms such as Bhashini and AI4Bharat, which support voice-first digital services in multiple Indian languages. Sector-specific uses include AI-enabled thermal imaging in health care in southern states and real-time landslide alerts in Himalayan regions.
“These solutions demonstrate locally grounded ingenuity,” the survey said, adding that SLMs allow rapid customization for regional languages and environments.
Data Governance, Jobs Shape India’s AI Vision
The survey also addressed concerns about employment and data control as AI adoption accelerates globally.
While large AI systems in advanced economies have raised fears of white-collar job displacement, India’s strategy seeks to “augment human value rather than replace it,” the document said. The focus is on using AI to enhance productivity and create higher-skill roles.
A proposed AI Economic Council will help calibrate the pace of adoption to ensure technology remains aligned with social stability and human welfare, the survey said. Officials believe sector-specific AI can help India evolve from an “IT back office” to an “AI front office.”
On data, the survey described information as the “core factor of production” in the AI era. India’s proposed governance framework favors accountable data portability over strict localization, allowing cross-border flows while requiring mirrored copies and local model tuning so economic value remains in the country.
The approach, the survey said, is designed to balance innovation, sovereignty, and inclusion as India scales its AI ambitions.
