India's AI Bottleneck Is Talent and Unusable Data, Not Big Tech
A survey of 308 stakeholders by The Dialogue finds talent shortages and a lack of data standardisation, not market concentration, are the binding constraints on India's AI ecosystem.
When Indian AI startups are asked what's holding them back, they don't say big tech. They say they can't hire, and they can't use the data they have.
That's the headline finding of 'Competition, Innovation, and Market Structure in India's AI Ecosystem', a report from policy think tank The Dialogue released this week. It surveyed 308 stakeholders — startups, business users and consumers — and found the constraints sitting well upstream of anything a competition regulator would recognise.
Talent is the joint top barrier
47.2 per cent of respondents named talent availability as a significant barrier to competitiveness, tying it with customer adoption for the top spot.
The nuance matters. India does not lack technologists. What it lacks, per Bhoomika Agarwal of The Dialogue, is the specialised layer above them — the skills to build, train, deploy and scale AI systems rather than consume them. A large services workforce and an AI engineering bench are not the same asset, and one does not automatically become the other.
The data problem is usability, not scarcity
This is the finding to sit with. 63.2 per cent of AI developers named a lack of data standardisation as their single biggest challenge. Copyright concerns came in at 44 per cent, and outright data scarcity at 42 per cent.
India, in other words, has data. It cannot get it into a shape that trains anything.
And the public repositories are largely shut: only 15 per cent of respondents said government-held datasets were readily accessible. For a country whose digital public infrastructure is held up as a global template, that number is a genuinely uncomfortable one.
Open source is load-bearing, not optional
96.2 per cent of AI startups and developers surveyed reported some reliance on open-source models and tools. 62.3 per cent described that reliance as high.
That is not a preference. That is the foundation the entire ecosystem is standing on, and it is a foundation maintained almost entirely outside India.
Why the framing matters
The report positions the sector as being in a rapid developmental and expansionary phase, with the structural deficits emerging as it scales rather than as evidence of capture by incumbents.
That framing has policy consequences. If the binding constraint were market concentration, the instrument would be competition law. If it's talent pipelines, data standards and access to state-held datasets, then the work is slower, less headline-friendly, and sits with education regulators, standards bodies and the departments holding the data.
The uncomfortable version: what is blocking Indian AI is the sort of thing nobody gets to announce at a summit.