(This interview has been lightly edited)

When you look at the global AI race today, what do you think are India’s deepest structural advantages? How much of that comes from our experience building digital public infrastructure and operating at population scale?

Whenever we look at technology for social transformation or for building impactful solutions and services, India’s core strength has been talent. India’s core strength has also been the ability to provide IT services to the whole world. From the Y2K crisis onwards, our major tech companies like TCS, Infosys, Wipro, Tech Mahindra and many others have been leading IT service providers globally. Indian engineering talent has been part of almost every big technology company. If you name any major company, there will be a significant number of people who studied in India, grew up in India and are part of that workforce. In fact, some of them are even leading these companies, such as Google and Microsoft. So we are known as technology service providers. If you look at the Stanford AI Index Report 2025, India ranks 2nd globally in relative AI skill penetration, just behind the United States. India is the second-largest contributor to AI projects on GitHub. Within India, the way we use technology is also important. The kind of digital public infrastructure we have built, whether it is Aadhaar, UPI or DigiLocker, that has again been a core strength. Our prime strength comes from human talent and from our ability to design and implement technology projects. When it comes to AI, that reveals our core strength. We have already shown how technology can be used for social good and for transforming services in both the public and private sectors. Coupled with the talent we have, that becomes our core strength. I don’t see it as a race. I see it more as every country trying to build its own capabilities. But if you look at how countries are currently stacked up, India is a bit behind the US and China. With current policy support and the initiatives that are ongoing, we do believe that we will be able to catch up with the best in the world.

Do you see an “India model” of AI infrastructure emerging that is different from models elsewhere in the world, where large companies often control everything from AI models to cloud infrastructure to end-user platforms?

When you talk about AI infrastructure, it primarily means compute, data centres, energy and chips. When it comes to chips, one company currently controls a large part of the market: NVIDIA. That is why it has such a large market value. But when it comes to the other components of the infrastructure like data centres, energy, power, renewable energy, there are multiple players. It is not that one company controls everything. In India, when we thought of building compute infrastructure, we felt that since it is very cost intensive, we should incentivise the private sector to invest in it. So we created a model where the private sector invests in compute infrastructure and we incentivise the end users. Through this model we were able to set up almost 40,000 GPUs. This is a very innovative model of financing compute infrastructure. Many countries that joined us at the AI Impact Summit in New Delhi last month were interested in this model. In fact, in the working group on democratising AI resources, a similar platform has been proposed so that a shared compute infrastructure can support countries in the Global South. So yes, the India model is different, and many countries are interested in replicating it. At the same time, it is not true that one company controls all layers. For example, NVIDIA designs chips but depends on Taiwan Semiconductor Manufacturing Company (TSMC) to manufacture them, and TSMC in turn depends on (Dutch semiconductor company) ASML for lithography machines. It is a very complex supply chain in which no single company or country has complete control.

India’s linguistic diversity and social complexity are often described as challenges. Do you see them instead as forces shaping a uniquely strong AI ecosystem, especially in areas like language AI, voice interfaces and inclusive access?

Yes, linguistic diversity is a challenge. When you build AI models or applications, you need datasets. Without datasets, models cannot perform as accurately as you would want them to. Take Wikipedia as an example. The English version has more than 7 million articles, while Hindi has only around 1.5–2 lakh articles. But if you look at languages such as Malayalam, Konkani, Dogri, Tamil or Bengali, the number of pages can fall to only a few thousand. So datasets in Indian languages are limited, and that limits the ability to train models that work well across multiple linguistic prompts. This is why, when we started the Bhashini project, we focused on creating datasets in all languages. We launched a crowdsourcing initiative called Bhashadaan, where native speakers contributed language datasets so that a corpus could be created. Once you have that corpus, you can train models and build applications. So it is a challenge, but it is also an opportunity. Through initiatives like Bhashini and models like Sarvam’s work on AI for Bharat, models trained from scratch in Indian languages and contexts are now performing better than OpenAI and Gemini on certain multilingual benchmarks. This approach can be very useful for countries in the Global South. Countries in Africa or Southeast Asia also face similar challenges with limited datasets in their languages. They can adopt a similar model. While linguistic diversity is a challenge, if we address it, the opportunity for building applications in multiple languages and voice-enabled services is enormous.

As India builds its AI ecosystem across digital public goods, talent, data and compute, how do partnerships with countries like Switzerland contribute to India’s long-term competitiveness?

India and Switzerland have been working together for a long time across various forums. Switzerland is part of the Global Partnership on AI, and India chaired it in 2023. We have worked closely with their teams. Switzerland was also very engaged during the AI Impact Summit last month in New Delhi. They supported several initiatives, including the Global AI Impact Platform, which is already live and contains more than 80 use cases from 30 countries. We also received support from Switzerland on voluntary guiding principles for resilient, innovative and efficient AI, which focus on sustainable AI and reducing energy consumption. Switzerland also endorsed guiding principles for reskilling in the age of AI and supported initiatives such as Trusted AI Commons and the Alliance for Advancing Inclusion in AI. Overall, Switzerland was very engaged during the entire summit. The objectives of initiatives such as the International Computing and AI Network are also aligned with the IndiaAI Mission, especially around democratising compute resources and building solutions in sectors like healthcare and education.

Looking back at the AI Impact Summit, what were some of the most important insights that emerged about how governments, industry and research communities can collaborate to ensure AI delivers meaningful social impact?

The summit took place from February 16 to 21. It brought together over 5 lakh participants from governments, CEOs, startups and civil society. There were more than 500 sessions across seven themes. One major achievement was the declaration endorsed by 92 countries, including the US, UK, Russia, China, the EU and Australia. Bringing together such diverse countries in the current geopolitical environment was a major achievement. The preparation for the Summit began months earlier. After the previous summit in France, we began preparations in July. What we did differently was make the agenda-setting process very inclusive and participative. We organised pre-summit consultations and events. Before the summit itself, we had already conducted around 600 events: about 500 in India and more than 100 globally. We consulted governments, industry, civil society and academia. That helped us curate the agenda for the summit. This inclusive approach helped ensure that countries felt ownership of the process. That is one reason why we were able to get 92 countries to sign the declaration and secure commitments from major AI labs like OpenAI, Google, Anthropic, Microsoft, TCS and Cohere. The key insight is that AI is a general-purpose technology. It cuts across sectors. It cannot be driven only by governments or only by industry or academia. It requires collaboration across all stakeholders, including civil society. Our Prime Minister also outlined a vision that emphasises innovation but also focuses on ensuring AI remains moral, ethical, safe, trustworthy and responsible. The idea is to build a framework where AI benefits everyone while reducing potential harm.

As we look ahead to the next global AI summit planned in Geneva, how do you see India and Switzerland strengthening their partnership?

We are looking forward to working with Switzerland to advance the commitments made at the New Delhi summit. For example, one initiative is the creation of a shared global compute infrastructure, which will require funding and cooperation from institutions such as the United Nations, the World Bank and the Asian Development Bank. I hope that by the AI Impact Summit in Geneva we will be able to come out with a framework for this. We also want to promote the adoption of AI applications in sectors such as healthcare and education across the world. Partnerships between academic institutions could help ensure that researchers and startups in the Global South gain access to compute, datasets and algorithms. We can also collaborate on capacity building for governments and regulatory agencies like sharing knowledge and best practices on AI governance. Geneva has a strong convening power because of the presence of the United Nations and many global institutions. Together we can ensure that the AI Impact Summit in Geneva reinforces the commitments made in New Delhi and leads to tangible outcomes.