India's AI Impact Summit drew more than 250,000 registered attendees and over $250 billion in combined investment pledges to New Delhi last week. The six-day event at Bharat Mandapam attracted CEOs from Google, OpenAI, Anthropic and Microsoft alongside 20 heads of state and 60 ministers from over 100 countries. It concluded with 88 nations endorsing the New Delhi Declaration on AI Impact. The numbers matter less for their size than for what they signal. India is attempting to rewrite the default assumption that artificial intelligence infrastructure will be built and controlled by a small number of Western and Chinese firms.
The summit was the fourth in a series of global AI gatherings that began with the Bletchley Park AI Safety Summit in 2023 and continued through Seoul in 2024 and Paris in 2025. It was the first hosted by a country in the Global South. The shift in branding from "safety" to "impact" reflected a deliberate pivot from theoretical risk frameworks toward deployment and measurable outcomes.
Infrastructure Commitments Take Center Stage
The headline figures came from Indian conglomerates. Reliance Industries announced plans to invest approximately $110 billion over seven years to expand AI infrastructure and services across India. Adani Group pledged $100 billion to build AI data centers powered by renewable energy by 2035. These domestic pledges were complemented by commitments from American technology companies. Microsoft confirmed it is on track to invest $50 billion by the end of the decade to expand AI capabilities in Global South countries. Google announced a $15 billion AI hub in Visakhapatnam along with its America-India Connect subsea cable initiative. OpenAI partnered with Tata Group to secure 100 megawatts of AI-ready data center capacity through Tata Consultancy Services' HyperVault business with plans to scale to one gigawatt. Anthropic opened its first Indian office in Bengaluru and announced an enterprise partnership with Infosys. Blackstone participated in a $600 million equity raise for Indian AI infrastructure firm Neysa.
The Indian government added to the momentum with a $1.1 billion allocation for a state-backed venture fund targeting AI and advanced manufacturing startups. It also announced the addition of 20,000 GPUs to the existing 38,000 under the IndiaAI Compute Portal. India announced plans to make compute available through this portal at approximately 65 rupees per hour, according to government officials, representing a significant reduction from standard market rates.
Sovereign AI Models and the Multilingual Challenge
Perhaps the most strategically significant announcements involved Indian-built AI models. Bengaluru-based Sarvam AI launched large language models with 30 billion and 105 billion parameters using mixture-of-experts architectures trained on domestic compute infrastructure. The company also unveiled AI-powered smart glasses called Sarvam Kaze and announced edge AI partnerships with Qualcomm and HMD to bring conversational AI to feature phones in Indian languages. The government-backed BharatGen Param2 model arrived as a 17-billion parameter multilingual system supporting 22 Indian languages. Gnani.ai introduced a voice model operating across 12 Indian languages under low-bandwidth conditions.
These launches address a structural gap in global AI development. Most large language models are optimized primarily for English, which limits their utility in a country where the vast majority of the population communicates in regional languages. India's approach to sovereign AI focuses less on competing with frontier labs on benchmark performance and more on building linguistically diverse systems designed for local deployment at lower cost.
Organizational Failures and Credibility Gaps
The summit's ambitions were undercut by significant execution problems. Multiple reports from attendees and journalists described overcrowding and long queues in the early days of the event. Bloomberg reported that delegates were left stranded without food or water during a security lockdown ahead of the Prime Minister's visit on February 19. The summit venue was closed to the public that day for the official inauguration, frustrating registered attendees who had planned to participate.
A notable controversy erupted when Galgotias University presented a robot dog at its exhibition pavilion that social media users identified as the Unitree Go2, a commercially available product manufactured by a Chinese company. The university was directed to vacate its stall. The incident drew particular attention because it undermined the summit's narrative of indigenous innovation. Bill Gates withdrew from a scheduled keynote address amid public backlash related to his past association with Jeffrey Epstein. Amnesty International published a statement arguing that the summit failed to address destructive AI practices by governments and technology companies. TechPolicy.Press criticized the summit's structure for granting corporations parity with sovereign governments through the CEO Roundtable while providing no equivalent platform for civil society or labor representatives.
The investment pledges themselves warrant scrutiny. The $250 billion figure represents commitments spread across many years. Reliance's $110 billion is planned over seven years. Adani's $100 billion target extends to 2035. Converting announcements at summits into operational infrastructure has historically proven more difficult than making the announcements themselves. India's existing GPU base of 38,000 units remains modest compared to the compute resources deployed by major hyperscalers.
What Enterprise Leaders Should Watch
For technology decision-makers outside India, three developments from the summit deserve sustained attention. First, the OpenAI-Tata and Anthropic-Infosys partnerships represent a new model for how frontier AI companies enter emerging markets. Rather than building their own infrastructure, they are partnering with established local conglomerates that already possess enterprise relationships and data center capabilities. This approach reduces capital risk while creating dependencies that could shape competitive dynamics for years.
Second, India's sovereign AI model strategy offers a potential template for other large developing economies seeking alternatives to full reliance on American or Chinese AI systems. The emphasis on multilingual capability and low-cost inference rather than frontier benchmarks reflects a different theory of what AI needs to do in markets where over a billion people speak dozens of languages and access technology primarily through mobile devices.
Third, the New Delhi Declaration, endorsed by 88 nations, represents the broadest multilateral AI framework to date, though it remains voluntary and lacks enforcement mechanisms. The White House delegation explicitly rejected binding global AI governance at the event, favouring market-led frameworks instead. Whether India can translate diplomatic momentum into durable governance norms will depend on follow-through at the next global AI summit scheduled for 2027. The gap between the summit's stated ambitions and its execution challenges offers an honest preview of the hard work that remains.