India AI Impact Summit Ends With New Delhi Declaration Backed by 92 Parties

The India AI Impact Summit 2026, the fourth in the series that began at Bletchley Park in 2023 and continued in Seoul and Paris, ran in New Delhi from 16 to 21 February 2026. The Indian government reported delegations from more than 100 countries and 20 international organisations, and an in-person attendance of roughly 600,000 across the week's public and expert programmes. The summit concluded with the adoption of the New Delhi Declaration, which the Ministry of External Affairs said was endorsed by 92 countries and international bodies, and which centres on inclusive, sustainable and human-centred AI, with an emphasis on developing-country access to compute, data and talent.

Investment announcements dominated the headlines. Government figures put expected commitments at more than 200 billion dollars across infrastructure, foundation models, hardware and applications. Reliance Industries pledged 110 billion dollars over seven years for AI infrastructure and Adani Enterprises 100 billion dollars by 2035; Google confirmed a 15 billion dollar AI hub in Visakhapatnam with new subsea cable routes; venture firms General Catalyst and Lightspeed announced 5 billion and 10 billion dollar programmes respectively. The IndiaAI Mission, which already provides more than 38,000 GPUs on subsidised terms, is to add 20,000 more. A "Global AI Impact Commons" was launched to catalogue more than 80 deployment case studies from over 30 countries.

The shift in framing is the story. Bletchley focused on frontier-model safety, Paris on investment and open innovation, and New Delhi on impact and inclusion, with safety present but no longer the organising principle. The declaration is non-binding, as its predecessors were, and no new international institution or verification mechanism emerged. What did emerge is the arrival of India as a compute and talent hub with state-backed GPU capacity, sovereign-model ambitions and a regulatory stance that favours guidelines over statute, offering a third model between EU rulemaking and US federal deregulation.

For multinationals, the practical consequences are about where capacity and workloads will sit. Large private and public GPU investments in India, combined with data-localisation expectations under the country's data protection law, will shape offshoring, delivery-centre and sovereignty decisions over the next several years.

What it means for leaders

• Reassess sovereignty and localisation strategies. Subsidised compute and new subsea routes make India a credible location for AI workloads; review data-transfer terms under the Digital Personal Data Protection Act before moving them.
• Do not read non-binding declarations as compliance relief. The New Delhi text creates no obligations; binding duties still come from the EU AI Act, national laws and sector regulators.
• Track divergence between summit tracks and hard law. With safety de-emphasised at summit level, frontier-risk controls will be driven by national regulators and by frameworks such as the NIST AI RMF and ISO/IEC 42001 that customers request.
• Use impact case libraries critically. Government-curated deployment stories are useful for scoping, but require your own risk assessment before replication.
• Prepare for procurement from Indian AI vendors. Rapid growth in domestic models and services will bring new suppliers into tenders; extend third-party AI due diligence accordingly.