India and the illusion of digital sovereignty: the technological asymmetry that questions Europe

The curtain came down on the Bharat Mandapam and the India AI Impact Summit 2026 came to a close: five days of announcements, group photographs between Altman, Amodei and Modi, and some inevitable organisational chaos. Among them, a robot presented as a product of Indian innovation and then discovered to be manufactured in China, with a consequent apology from Minister Vaishnaw. An embarrassment that is not just folklore, but the perfect synecdoche of the Indian problem: the casing is domestic, the technological engine is foreign. I am among the first to argue that India is an inescapable pivot of the global technological order. But what I observed in New Delhi is a more complex and alarming story.

G20 Sherpa Amitabh Kant stated that Indian users generate 33% more interactions than Americans on ChatGPT. The idea promoted: 'we are the reservoir of the future'. What OpenAI confirms is different: India has over 100 million weekly users, the largest market outside the US. The problem is that using ChatGPT is not the same as providing data to train the models of the future: large models are trained on assembled corpora before the user touches the keyboard. Confusing traffic and power in AI is a categorical error that suits the Indian political narrative. OpenAI's partnerships with Tata, Anthropic and Infosys confirm a structural asymmetry: Silicon Valley seeks markets, talent and political legitimacy; India seeks technologies it cannot yet build itself. This is not an alliance: it is an asymmetric partnership with a hidden price.

Reliance and Adani have announced $210 billion in data centres over the next decade. But where does this money go? It goes into 'iron': cables, servers, cooling, power. The core of value (frontier models, chips, alignment systems) remains an American oligopoly: OpenAI, Anthropic, Google DeepMind, Meta AI. The risk is to build magnificent highways for cars produced elsewhere to run on. This is not digital sovereignty: it is, at best, qualified participation in a system that others govern.

The most dangerous node is the social detonator hidden in the models that India is importing. American AI systems are productivity optimisers designed for economies with labour shortages and high wages. To bring these systems into an economy that adds seven to eight million new workers every year, where IT and business process outsourcing employ tens of millions of people in tasks of exactly the kind most susceptible to automation, is an unprecedented social experiment. India has the historical merit of having built a Third Way in digital payments (UPI system) by resisting American monopolies with a global model of financial inclusion. But generative AI is not a payments system: it requires proprietary computing power that New Delhi does not possess today. One imports the American paradigm of unbridled automation, one does not invent a new one. The Indian middle class was formed through the lift of working in IT. If that lift stops, the strains would exceed in magnitude those of European de-industrialisation, in a country with far less developed welfare systems.

Europeans risk being the most qualified, and most irrelevant, spectators in this game. The AI Act has reaffirmed the world's reputation as regulators. But an ethics committee without an engineering department has limited influence. The opportunity for Europe, and I say this as president of the Digital Economy Centre, exists and is concrete: our regulatory architecture, our credibility as a third party to the US-China competition could form, alongside industrial co-investments to create sovereign computing infrastructure, the basis of a triangular alliance with India. The geopolitics of AI in 2026 is not only played out over semiconductors and models. It is played out on the resilience of social contracts in the face of the automation wave, and on the ability of those who are neither Washington nor Beijing to still have a say. India has chosen to try. We are still working on it.