At a summit meant to showcase India’s growing clout in artificial intelligence, the loudest conversations this week were not only about new language models or computing capacity, but about who gets to build them, who gets to claim them, and how far the country can go without borrowing someone else’s machinery.
The India AI Impact Summit 2026, held from February 16 to 20 at Bharat Mandapam in New Delhi, drew heads of state, ministers, company founders, and hundreds of start-ups into a tightly choreographed five-day schedule that mixed research sessions with investor announcements and public exhibitions. Prime Minister Narendra Modi opened proceedings by touring stalls run by Indian firms such as Sarvam AI, HCLTech, and Reliance Jio, while urging developers to pursue responsible use of artificial intelligence in areas such as agriculture, health care, and education.
Yet even as the summit pushed the idea of home-grown models and language systems suited to India’s many scripts and speech patterns, a string of controversies exposed a deeper tension: the gap between national ambition and the hard reality of dependence on foreign hardware and software.
That gap came into sharp relief on the opening day when organisers were forced to ask representatives of Galgotias University to vacate their stall after they presented a Unitree Go2 robotic dog as an in-house creation. Online sleuths quickly identified the quadruped as a Chinese commercial product, leading to ridicule on social media and renewed debate about whether India’s drive for technological self-reliance risks being undermined by overstatement.
Union IT Minister Ashwini Vaishnaw later acknowledged logistical problems that plagued the first day, including long queues, patchy internet access that disrupted live demonstrations, and payment counters that accepted only cash at an event centred on artificial intelligence. A control room was set up to address complaints, and digital payments were restored from the second day onward. Some exhibitors also reported the loss of wearable devices after security sweeps cleared halls before the prime minister’s visit.
Despite the messy start, the summit moved quickly into policy discussions. Vaishnaw announced plans to add 20,000 graphics processing units to the country’s compute pool, supplementing the existing 38,000 already in use for research and commercial work. The move is intended to reduce reliance on overseas data centres that currently handle much of the training work for large-scale models.
India’s approach rests heavily on language diversity. BharatGen’s Param2 model, built by researchers at IIT Bombay, supports 22 Indian languages and is trained on data drawn from government archives, literature, and broadcast media. Sarvam AI introduced new products, including Bulbul V3 and Arya, aimed at speech recognition in regional languages, while the start-up Gnani showcased VoiceOS for conversational interfaces in customer service settings.
Panels across the second and third days examined artificial intelligence in manufacturing, farming, and public service delivery. A symposium on AI for social welfare brought together officials from UNICEF and education ministries to discuss classroom use, while sessions led by chipmaker Nvidia focused on building domestic semiconductor capacity over the next two decades. Discussions on defence uses of robotics and autonomous systems drew a packed audience.
Investment announcements also came thick and fast. Mukesh Ambani said Reliance Jio plans to spend ₹10 lakh crore over seven years on artificial intelligence research and data centres powered by renewable energy. The Adani Group has set out plans for AI-linked data centres worth $100 billion by 2035, while a state-backed venture capital fund of $1.1 billion for AI start-ups was announced on the sidelines.
Taken together, analysts estimate that as much as $200 billion could flow into India’s artificial intelligence industry within two years. For comparison, government spending on information technology research across all ministries in 2024–25 stood at roughly ₹18,000 crore, according to budget documents. The scale of private investment now being discussed suggests a rapid change in how the country funds advanced computing research.
International participation lent the summit added weight. Estonia’s President Alar Karis and French President Emmanuel Macron addressed plenary sessions on governance and free expression in the age of machine-generated content. Switzerland’s President Guy Parmelin expressed interest in hosting the next summit in Geneva in 2027.
Sir Demis Hassabis of Google DeepMind delivered a keynote address on research trends, while Sundar Pichai visited the exhibition halls alongside venture investor Vinod Khosla. Khosla warned that many jobs in India’s IT and business process outsourcing industry could disappear as companies turn to automated coding and customer support systems, prompting calls from officials for reskilling programmes.
One of the week’s lighter moments came when OpenAI’s Sam Altman and Anthropic’s Dario Amodei appeared to avoid physical contact during a group photo on stage, opting for an air fist-bump rather than joining hands with other executives. The image quickly circulated online as a symbol of the rivalry between companies competing to build general-purpose AI systems.
A more serious absence was that of Bill Gates, who cancelled his scheduled appearance at short notice amid renewed public scrutiny of his past association with Jeffrey Epstein. While organisers did not link his withdrawal directly to the controversy, international media reports framed it in those terms, shifting attention away from policy discussions.
For India’s government, the summit served as an opportunity to set out its “MANAV Vision” for human-centred governance of artificial intelligence. Officials argued that public data rails and open software stacks can help ensure that smaller towns and rural districts benefit from automated systems used in health care or crop management.
Examples on display included sign-language recognition software for classrooms, predictive models for soil moisture levels in dryland farming, and adaptive learning programmes for children with special needs. Representatives from Tata’s AI Sakhi immersion programme described efforts to train rural women to use AI-based applications for microfinance and crop planning.
Adobe announced free access to Photoshop, Acrobat, and Firefly for accredited educational institutions through a partnership with NASSCOM, along with online courses in machine learning basics. Semiconductor firm AMD said it will work with Tata Consultancy Services on computing capacity for research laboratories, while Qualcomm outlined plans for on-device AI in smartphones.
These announcements fed into a wider debate about whether India can match the United States and China in building large-scale computing capacity. While domestic companies can write code and train models, much of the specialised hardware used in training still comes from abroad. That reliance has drawn attention from policymakers concerned about supply disruptions or export controls.
India’s semiconductor roadmap, discussed at several sessions, seeks to build fabrication plants within the country over the next twenty years. Officials said the aim is to cut dependence on imports while supporting research in natural language processing for Indian languages.
Attendance figures suggest strong public interest. Organisers estimate that more than 150,000 visitors passed through the exhibition halls during the week, with the expo extended to February 21 due to demand. Over 2.5 lakh students also signed a pledge committing to responsible use of artificial intelligence in research and business.
Still, the summit’s first-day confusion, the robotic dog incident, and the absence of a high-profile speaker exposed weaknesses in planning and messaging. Opposition leaders seized on the early chaos as a sign of administrative lapses, while supporters argued that such problems are inevitable at events of this scale.
