India used the India AI Impact Summit 2026 in New Delhi to present itself as a convenor on artificial intelligence at a time when governments are still arguing over what “safe”, “open” and “sovereign” AI means in practice.
The event ran from 16 to 20 February 2026 at Bharat Mandapam (Pragati Maidan), combining a high-level summit with an expo and a steady flow of ministerial statements and corporate messaging.
For European policymakers, the summit was also a diplomatic fixture. The European Commission said Henna Virkkunen, the Commission Executive Vice-President responsible for technological sovereignty, security and democracy, would endorse a Leaders’ Declaration on the EU’s behalf. In parallel, India’s foreign minister S. Jaishankar marked the launch of a European Legal Gateway Office in India, framed as a practical step to connect European companies with Indian capability and to deepen the EU–India strategic partnership.
The most eye-catching claims were about infrastructure. India’s IT minister Ashwini Vaishnaw said that more than 38,000 GPUs had been onboarded under the IndiaAI mission, and that 20,000 additional GPUs would be added. A separate report cited a price point of around Rs 65 per hour for access, presented as a way of lowering entry costs for startups and researchers. He also spoke of investment pledges in the hundreds of billions of dollars, with some Indian media reporting figures around $270bn.
Those numbers are politically useful, but they also raise basic questions that remain largely unaddressed in public read-outs: what counts as a “pledge”, how much is new money rather than repackaged pipelines, what time horizon is being assumed, and how far the build-out depends on imported hardware and foreign cloud capacity. Even supportive international coverage has noted the tension between India’s sovereignty rhetoric and the practical reliance on overseas semiconductors and external data infrastructure.
That tension mattered because the summit’s language repeatedly returned to “sovereignty”. The United States used the gathering to promote what it described as “sovereign” AI for allies alongside adoption and exports, which, in plain terms, is also a pitch for American systems, standards and supply chains. If India’s fastest route to scale runs through US model providers and US-aligned stacks, the question is whether “sovereign” becomes a branding term rather than an industrial reality.
The governance discussion was similarly broad-brush. The Leaders’ Declaration, as characterised in official summaries, leans on shared benefits and high-level commitments rather than enforceable obligations. India’s own official communication described the summit as symbolism: a “Global South” venue for a “defining global technology conversation”. Symbolism is not nothing, but it does not settle disputes over liability, auditability, redress, and the use of AI in state functions.
Internationally, the fault lines were visible. Reporting from the summit noted a UN push for “science-led governance” and “human control”, while the US delegation warned against centralised control of generative AI, underlining the difficulty of landing anything more than a lowest-common-denominator communiqué. This is the structural problem for any leaders’ declaration: it can show alignment on aspiration while papering over the regulatory mechanics that determine how systems are built and deployed.
From an EU perspective, the summit sits awkwardly alongside Europe’s own regulatory direction. The Commission’s public line in New Delhi emphasised cooperation, innovation and international governance, while announcing initiatives such as a “Frontier AI Grand Challenge”. Yet Europe’s credibility in global AI talks will increasingly be tested on whether it can combine regulation with competitive scale, and whether cross-border partnerships do more than create talent pipelines for non-European model ecosystems.
There were also questions left in the background about energy and skills. One report from the summit carried Microsoft president Brad Smith’s message that infrastructure alone is insufficient without large-scale training and capability-building. That point is obvious, but it matters because “compute as a public good” only works if the public also has the capacity to use it, and if the electricity and connectivity required for data centres do not become a bottleneck.
India’s summit therefore achieved what such events usually achieve: it gathered political attention, offered a platform for announcements, and helped align a narrative around national ambition. What it did not clearly provide—at least in the public documentation so far—is the harder material: precise definitions of sovereignty, transparent accounting of investment claims, and a concrete framework for accountability when AI systems cause harm or are used for coercive ends. The summit may still influence the agenda. The test will be whether the next steps move beyond declarations into verifiable policy and measurable outcomes.
