India’s AI road dotted with challenges

INDIA has every reason to be proud of the success of the AI Impact Summit. This was the fourth in a series of summits which began at Bletchley Park (UK) in November 2023, followed by Seoul in May 2024 and Paris in February 2025. While the focus of the first two summits was on AI safety and risk management, the Paris and Delhi summits shifted the emphasis decidedly towards a more positive and forward-looking narrative.

If the Paris conference billed itself as the AI Action Summit, the Delhi meeting went further to brand itself as the AI Impact Summit. The Paris event was co-hosted by French President Emmanuel Macron and Prime Minister Narendra Modi, while the Delhi summit had Macron as a principal guest who delivered the keynote address. This was solely India’s show, marking the first instance of a country from the Global South hosting such a summit.

While there were several heads of state/governments in attendance, the AI superpowers — the US and China — were decidedly under-represented. The US delegation was led by Michael Kratsios, Assistant to the President and Director of White House Office of Science and Technology. He is said to be the AI czar in the Trump administration. China was represented by its Vice Minister of Science and Technology, Chen Jiacheng. In Paris, the US was represented by Vice President JD Vance and China by Vice Premier Zhang Guoqing.

However, the US was over-represented by its high-tech industry giants, including Sam Altman of OpenAI, Dario Amodei of Anthropic, Demis Hassabis of Google DeepMind and Sundar Pichai of Google/Alphabet.

The summit was part policy forum, part technology and trade exposition, with countries as well as companies showcasing their capabilities and products. For the large number of Indians visiting the sprawling venue of Bharat Mandapam, this was a rare and welcome exposure to frontier technologies and their promise for economic and social development. It was also an opportunity for Indian companies to display their mastery of high technology, though the unfortunate instance of Galgotias University passing off a Chinese-built robot as its own was a disastrous public relations snafu.

While the Paris Declaration (2025) was rejected by the US and the UK, reportedly because it emphasised global and inclusive governance and neglected national security concerns, the New Delhi Declaration was a consensus document endorsed by the UK and the US. This required some creative drafting. The declaration emphasised “democratic diffusion” rather than inclusivity. The reference to Global AI Impact Commons suggested the need for norms but avoided any reference to multilateral norm-setting or global governance.

The United Nations was ignored. This may have kept the US on board, but may pose a challenge for India going forward. For a developing country like India, multilateralism and global governance are indispensable principles in dealing with challenges which are cross-border, cross-domain and truly global in dimension. AI is precisely such a challenge. It is a general-purpose technology of immense power and promise. Its domination by just one or two nations will limit the agency of developing countries.

A visit to various stalls at the summit was inspiring in many ways. One witnessed many creative applications of AI geared to India’s specific developmental needs. They offered solutions at the grassroots level and in a form and language that could be tailored to the communities being served. There was an AI-assisted optical scan which could provide almost instant data on several vital health indicators. Or AI-enabled tablets which can store all textbooks that a child takes to school and enable learning through text and visuals.

A demonstration of the use of AI in healthcare and education was the most impressive. It was also inspiring to see young Indians explaining complex technologies and applications with confidence and enthusiasm. If our older generation is fearful of the significant risks that AI clearly entails, this generation is excited by its promise and embraces it boldly.

There is no doubt that the summit established India as a significant and aspiring AI power. While PM Modi did sound a cautionary note about the importance of safety, the emphasis was clearly on making India one of the leading AI powers. The summit presented this goal as an exciting and rewarding forward trajectory. But the challenges and constraints are sobering.

AI depends upon immense data-crunching capacity which comes from high-end Graphics Processing Units (GPUs). The US is estimated to have 1.5 million GPUs in operation, with a target of three million over the next few years. China has half a million GPUs, with a target of 800,000. India has only 38,000, with a target of 100,000 by the end of this year. The whole of Africa has only 1,000 GPUs.

PM Modi announced at the summit that India aspired to become the data centre for the world. But data centres require a huge scale of power and large quantities of water for cooling. A 100-megawatt (MW) AI data centre can use up to 1.9 million litres of water per day — roughly what a small city consumes.

Currently, India has 1.5-1.7 gigawatts (GW) capacity of data centres, with a target of 8 GW by 2030. The corresponding capacity of China and the US is 7.5-9 GW and 14-16 GW, respectively. Their target is 12 GW (China) and 25 GW (US) by 2030.

Given the scale and variability of power consumption, grid power is not always the best source of electricity. Most data centres are settling for captive power and exploring small modular nuclear reactors as the solution. For India, an energy-poor country, with electricity consumption per capita lower than the global average, AI plans must not be at the cost of efforts to reduce and eliminate energy poverty. Diversion of power for catering to ambitious AI plans would be counter-productive.

The same argument applies to water availability. Water security is already a growing challenge and the volume of water required for even a modest AI plan may expand economic inequalities and spawn undesirable social and political consequences. India’s AI ambitions must be tempered by these inescapable realities and should be harnessed to removing these constraints.