India should be realistic about AI sovereignty. Given where it is, 50% of the job done will be creditable 

The year opened with Trump capturing Maduro, underscoring the extent to which even a major state’s sovereignty can be constrained by US military reach. When it comes to AI, however, almost every nation except US is similarly constrained – able to develop and deploy AI only so far as global tech supply chains, cloud infra and chip exports permit.

Consider China. At the start of last year, its DeepSeek AI models were disrupting US markets by outpacing established chatbots and prompting stock gyrations. This month, China’s ByteDance released Seedance 2.0, an advanced text-to-video generation model that has gone viral domestically and caught global attention for its film-quality outputs. Hollywood studios and rights holders have already raised copyright concerns, arguing that such models operate without meaningful safeguards against infringement. Despite the capability on display, China’s AI ecosystem remains only partly sovereign – heavily dependent on foreign-made high-end chips that Washington can restrict.

India is even further from AI sovereignty. It doesn’t yet possess a homegrown foundational model of anywhere near world-class scale, nor does it have the sprawling, hyperscale data-centre capacity needed for large-model training and inference at scale. That’s why the Adani Group’s announcement of a planned $100bn investment in renewable-powered AI data centres and a sovereign cloud platform by 2035 is significant, even if stretched over a decade. This commitment, unveiled at the AI Impact Summit in Delhi, comes amid wider interest in India’s AI infra build-out: govt is reportedly targeting up to $200bn in data-centre investments to make India a global AI hub. Already, Adani has explored investing in Google’s AI data-centre projects in India, signalling private-sector appetite to add compute muscle.

Meanwhile, worldwide AI development is exploding. Computation used for training powerful models – often described as “training compute” – continues to rise rapidly, demanding vast clusters of specialised chips. Large tech firms have poured vast sums into AI: US alone saw nine-figure investments in 2024 and continued heavy spending in 2025 and 2026, dwarfing early Indian commitments.

This intensity reflects a simple truth: AI is no longer a novelty. Tools like Claude’s latest upgrade are reshaping professional work, while video models such as Seedance now unsettle creatives. India’s advantage – a huge talent pool – gives it one pillar of strength. But without serious investments in compute and foundational model research, it will only ever achieve “half-sovereignty.” Chips, meanwhile, remain a longer-term endeavour. Even half sovereignty, however, is better than none – and India’s current moment is an opportunity to build it.

END OF ARTICLE