India should prioritise AI applications, not large-scale frontier models, says Economic Survey

India’s competitive advantage in the artificial intelligence era lies in application-led innovation rather than attempting to replicate frontier-scale model development, the Economic Survey released on January 29 stated. While efforts to build large foundational models can generate valuable process knowledge, the survey said India’s real strengths rest in its domestic data, human capital and the ability of public institutions to coordinate large-scale initiatives.

The observations come amid an ongoing debate within India’s technology and startup ecosystem on whether the country should prioritise foundational AI models or focus on the application layer. The survey argued that a narrow pursuit of scale for its own sake is less effective than a strategy rooted in practical use cases.

The document advocated a bottom-up approach to AI development anchored in open and interoperable systems, sector-specific models and shared physical and digital infrastructure. It stated that such a pathway offers a more credible route to value creation than focusing exclusively on large, compute-intensive models. According to the survey, application-specific and smaller models are significantly more computationally efficient, easier to fine-tune and capable of running on locally available hardware such as smartphones and personal computers, making them more compatible with India’s existing infrastructure.

The survey further noted that this approach allows innovation to emerge from a wider range of stakeholders, including startups, research institutions, public agencies and domain-specific firms. By lowering resource requirements and entry barriers, it enables broader development and diffusion of AI solutions across sectors without being constrained by access to massive computing power or capital.

Emphasising the role of open-source ecosystems, the survey stated that India should encourage innovation on open-source and open-weight platforms to achieve more with fewer resources. It highlighted that India is among the world’s largest and fastest-growing communities of open-source developers, many of whom actively contribute to global codebases in collaborative environments. The survey said unifying these efforts under the IndiaAI Mission would be essential to channel this talent towards shared domestic innovation.

Providing India’s talent pool with a robust platform and clear policy guidance could reduce dependence on foreign proprietary systems, lower entry barriers for domestic developers and foster an environment that encourages experimentation across sectors at relatively low cost, the survey stated.

To coordinate this bottom-up strategy, the document proposed an ‘AI-OS’ initiative, drawing parallels with public digital infrastructure such as UPI and Aadhaar, with the aim of positioning AI as a public good. The survey said such an initiative would enable the sovereign to collaborate with state and local institutions to expand access to structured, anonymised and machine-readable datasets in priority sectors, pool existing data centre capacity to create shared cloud computing infrastructure, and establish common platforms where open-source AI efforts can be coordinated and audited.

In contrast, the survey noted that sector leaders in Western economies have largely pursued a top-down approach centred on frontier models, massive private capital investment, heavy spending on computing infrastructure and the concentration of intellectual property within a small number of hyperscale firms.