At last count, half a dozen Indian states have promulgated AI policies and over a dozen have announced AI Centres of Excellence. At the national level, the ₹10,000+ crore IndiaAI Mission has put money and political attention behind the ambition to make AI in India and AI work for India. These are welcome steps, and show seriousness about a subject that barely registered three years ago.
This first chapter of India’s AI story has been about arrival: Attracting investments and signalling seriousness on the global stage. The second chapter is about foundation: building the market and institutional substrate. It is a difference between if India becomes an economic beneficiary of AI or, as former Reserves Bank of India governor, Raghuram Rajan, argues, merely a tenant--providing land, power, and labour while the value resides elsewhere.
AI economy rests on six interdependent layers: Energy, chips, infrastructure, data, models, and applications. Only a handful of countries will lead across all six. For the rest, the rational response is to build depth where comparative advantage exists and forge alliances where it does not. India’s acknowledged strengths lie in data volume and application development.
However, knowing your strengths is not the same as having a strategy.
{{/usCountry}}However, knowing your strengths is not the same as having a strategy.
{{/usCountry}}Development economics offers a sobering corrective. In the 80’s, the Washington Consensus prescribed ten market-oriented reforms to the developing world. Several countries that adopted it stagnated. Several others that ignored it industrialised. It failed by confusing conditions for success with a path to them.
The AI era is already generating its own version of the Washington Consensus: Install megawatts, build data centres, reskill workers, publish an AI policy, attract the hyperscalers, and wait for transformation to follow. This best-practice playbook risks the same mistake: treating necessary conditions as sufficient for success.
Growth is delivered through the invisible, often painstaking, creation of a flywheel that transforms inputs into outcomes. States that don’t build this unglamorous flywheel risk being left behind.
A robustly documented insight in industrial development is that the most consequential lever a government holds is not its budget but what it buys.
The US did not build the internet, GPS, or the semiconductor industry solely through market competition. It built them because the department of defence was a demanding, technically sophisticated early customer--creating the anchor demand around which private innovation organised itself.
Anchor demand for AI applications is the single most powerful thing Indian states can do to stimulate local innovation. India Stack is, in principle, one of the most powerful AI-enablement platforms in the world, but a platform is not a market. UPI succeeded because NPCI actively catalysed adoption. A PLI-like incentive for private developers building AI applications on the Stack is what is needed. Enough to create a commercially viable cohort, after which market logic can sustain what policy initiated.
AI is only as useful as the data on which it operates. 1.4 billion Indians, amongst the most technology-friendly citizens globally, are generating data at an enormous pace. Yet most Indian states are haemorrhaging their most valuable data asset without appearing to notice. Agricultural transactions, land records, health histories – this data either flows silently to foreign servers or sits in unusable paper form in locked almirahs.
Estonia’s lesson is prosaic: In the 1990s, that small nation neither attracted technology companies nor built data centers. It spent a decade digitising its entire public record across land, identity, health, and business registration; making it interoperable across government. Today, Estonians access majority of public services digitally.
India’s data challenge is two distinct problems requiring two distinct responses. The first is a problem of flow: Data being generated today in government settings that is not being retained by the State. This data is either produced on paper or generated digitally in formats that reside on vendor servers. Mandating State-owned digital record is a governance decision requiring changes to procurement standards, vendor contracts, and departmental procedures. None of these are photogenic. All of them are foundational.
The second is a problem of stock: Historical records sitting in paper form across district offices, sub-registrar rooms, and public hospital files. It is one of the most valuable and irreplaceable datasets a state government holds that data centre investment cannot replace.
AI is not, fundamentally, a technology problem that can be delegated to a specialist department. It cuts across fiscal policy, procurement rules, education curriculum, labor regulation, data governance, and public service delivery. No technology ministry or Centre for Excellence can coordinate across all of these. Believing the challenge has been addressed once a dedicated body is created may prove the most consequential error of this moment.
Singapore’s Smart Nation initiative is the most instructive example: it succeeded because the agenda was owned and driven directly from the Prime Minister’s Office, with the authority to compel integration across every ministry and override departmental inertia.
States that genuinely lead in the AI era will be those whose chief ministers treat AI not as a technology sector to regulate or a showcase to display, but as a governing subject to personally own. That means procurement reform driven from the top. Data policy that cuts across every department. The Centre of Excellence may be where the research takes place. The centre of government is where the architecture is built.
The data centre, the reskilling programme, the AI Centre of Excellence: These are visible, announceable, and real. The skeptic who dismisses them is wrong. But the source of power lies elsewhere. India has, with speed and seriousness, completed the first chapter of its AI story. The second chapter is harder, less visible, and more important. The question is whether its states have the institutional imagination and the political courage to write it.
(The views expressed are personal)
This article is authored by Vivek Agarwal, country director-India, Tony Blair Institute for Global Change.