India accounts for just about 2% of the startups globally working on AI training data, compared with 40% in the US, 21% in the EU, and 9% in the UK. China accounts for around 5%, the survey said, citing World Bank data.
“India holds considerable potential advantage in terms of domestic data sources, but this asset remains underutilised,” the survey stated.
Globally, AI development has split into two paths. Western tech leaders have focussed on building large frontier models backed by massive capital and computing power, while many other countries have adopted a decentralised, application-driven approach, the survey explained, noting that India should focus on the latter.
The report suggests that India should have a bottom-up approach to AI development, focussing on application-specific solutions rather than competing directly in building large, expensive foundational models.
Given limited access to advanced computing infrastructure and funding, India is better placed to focus on smaller, targeted AI models tailored to specific sectors such as healthcare, agriculture, finance, education, and governance, it said.
“These models are more efficient, easier to fine-tune, and better suited to India’s existing infrastructure,” the report noted.
Such an approach would allow startups, research institutions, public institutions, and domain-specific companies to play a larger role in innovation, lowering entry barriers and improving AI adoption across sectors.
The report also urged greater participation from Indian companies to scale successful AI applications and absorb early-stage risks.
It said that developing sector-specific AI solutions can help India’s IT industry climb up the value chain, moving from a back-office services provider to an ‘AI front office.’
Overall, the survey highlighted that a bottom-up, open, and application-focussed AI strategy offers the country a roadmap to build a globally competitive AI ecosystem.