India AI Impact Summit: Powering inclusion at global scale

India’s AI moment has arrived — and it will now be judged not by ambition, but by architecture. As the country prepares to host the India AI Impact Summit 2026, the focus shifts from startup buzz and research depth to compute muscle, data sovereignty, and deployment at population scale. With the Global South asserting its place in shaping AI governance, India’s execution model is under a global spotlight.

At the centre of this inflection point is Abhishek Singh, Additional Secretary, Ministry of Electronics and Information Technology, and CEO of the IndiaAI Mission. A 1995-batch IAS officer who has steered key national digital platforms including NeGD, MyGov and Karmayogi Bharat, he now anchors India’s AI strategy across emerging technologies, cybersecurity and digital skilling.

In this exclusive conversation with Voice&Data, he lays out the Summit’s strategic intent, the blueprint for democratised compute and datasets, the push for sovereign foundation models, and why true success will be measured by integrating 500 million more Indians into the digital economy through voice-led AI services. Excerpts:

How does the upcoming India AI Impact Summit fits into the AI Mission’s execution roadmap during the next two to three years, and what tangible outcomes can the stakeholders expect?

This Summit is extremely significant for us. It is the first such major AI Summit being held in the Global South, and it has been conceptualised with three key objectives.

The first objective is to position India as a country of strength when it comes to building AI solutions. India has long been regarded as a technology-strong nation—a country with tremendous engineering talent, with Indian companies building software and digital systems for multiple countries across the world.

We want to position India as the go-to place, both for sourcing AI talent and for outsourcing AI-related work to Indian companies. As we move forward, AI transformation services are going to be in great demand globally. Given the capability of Indian companies, startups, researchers and engineers to build impactful AI solutions, we want the world to look at India as a trusted partner in building those solutions.

Secondly, the AI agenda of the Global South is at already the forefront. What we want to demonstrate is a model that democratises AI—democratises access to datasets, democratises access to algorithms, democratises access to compute.

The IndiaAI Mission has enabled low-cost compute and access to datasets in a structured manner. We believe the platform we have created can offer examples for the Global South to replicate—especially countries that may not have the same financial resources but want inclusive AI development.

The working groups we set up for the Impact Summit have been working towards aligning these objectives. We are hopeful that the declaration adopted at the Summit will create a framework through which AI development and deployment will reflect a stronger voice of the world.

The third objective relates to startups and capital. India has a very strong startup ecosystem. But very often, many brilliant AI startups and brilliant minds move to the West because funding is more easily available there. So, we are trying to encourage the VC ecosystem and global investors to invest in India’s AI story—to ensure that we retain our talent, enable startups to access funding, create intellectual property, and contribute to national income.

The India AI Impact Summit will demonstrate how democratised compute, sovereign models, and shared datasets can power inclusive AI at global scale.

The response to the Summit has been overwhelming. Leaders from the global tech community, research community and policy community are participating. It reflects the confidence and promise the world sees in India’s AI journey. I am confident that we will be able to present to the world India’s way of implementing AI.

AI in India is now moving from policy intent to implementation. What are the top priorities in the 12 months following the Summit that will signal real progress rather than incremental announcements?

The top priorities are about capacity, scale and adoption. We need to build capacity at all levels. We need to expand the availability of compute infrastructure. We need to build data centre capacity so that India becomes the inference capital for AI workloads globally.

We must ensure that the foundation models we started funding last year—which are now maturing—are launched and adopted. Many of them are in advanced stages and will be launched during the Summit.

There is also a need to ensure adoption of these models across the country across various sectors. We must build impactful applications that transform healthcare, education, agriculture and other domains. We are also building the orchestration layer—so that inferencing can happen at scale; it will help anyone building AI applications leverage common infrastructure.

And very importantly, we must ensure that more and more people are able to benefit from AI-enabled services. Ultimately, we must create a strong talent pool in India—so we can take up the jobs emerging in the AI economy and the data economy in the days to come.

In the next 12 months, we should see adoption of AI applications, mushrooming of AI startups building impactful solutions, nurturing of talent, and strengthening of responsible AI at population scale.

As the Summit communication highlights, India is making efforts to redefine AI as “all inclusive”. Could you explain or give some impressions of what exactly will it change and offer the citizens?

Very often, when digital services are introduced in India, people who cannot read or write English are left out. Most online services also require accessing a browser, and even the URL is primarily in English. When you go to a website, even if it offers multiple languages, you may need to know English just to toggle into another language. Similarly, many people are not comfortable typing on a browser or mobile app. Even accessing a captcha becomes a hurdle.

India is a voice-first country. People speak in their local languages. They interact in their mother tongues. So, what they need are services that are voice-enabled and available in local languages. They should be able to access online services simply by giving voice prompts and receiving responses—without needing to type.

AI has immense potential here. Through initiatives like Bhashini, we have already enabled multilingual capabilities. As we integrate natural language processing (NLP) and generative AI into sectoral applications, more people will be able to connect to the digital world. And when that happens, their efficiency and productivity will improve and their incomes will go up. That would lead to higher per capita income and national income growth. This will contribute to the country’s journey from a USD 4 trillion economy towards a USD 30 trillion economy—and towards Viksit Bharat.

Compute access has emerged as a strategic bottleneck globally. What is India doing to ensuring affordable, reliable and scalable compute for startups, researchers and public sector deployments?

We are incentivising private sector investment in compute infrastructure, while subsidising end users. Today, researchers, students and startups get around 40% subsidy in accessing compute. Also, we came up with an innovative bidding and tendering process enabling private sector to build compute infrastructure. Today, GPU is available at an average cost of around Rs 65 per GPU per hour in India, compared to the global cost of USD 2 – 3 per GPU per hour.

This has enabled affordable access. We currently have around 38,000 GPUs. If we have to meet the requirements of startups, industry and especially if we have to do inferencing at scale across the country, we need to expand that capacity significantly.

So, the next step is expansion—both in scale and in reliability.

India’s Digital Public Infrastructure model has been globally recognised. Can a similar architectural approach be applied to AI—treating compute, models, datasets and tools as shared national assets rather than fragmented resources?

Yes. In fact, our AI strategy is based on the DPI story. Under DPI, we built the common identity layer—Aadhaar; the payments layer—UPI; the data layer—DigiLocker, the data architecture, and data sharing frameworks. On top of those common rails, applications were built.

Similarly, AI requires three things: compute, datasets and foundation models. Compute is expensive. Not everyone can invest in it. Datasets are critical and come from multiple sources. Foundation models are essential—otherwise you will remain dependent on foreign models and lack sovereign capability.

Foundation models, compute and datasets are means to an end; real success lies in bringing 500 million more Indians into the digital economy.

Under the IndiaAI Mission, we have created a common compute layer available at affordable cost. We have created AI Kosh—a datasets platform—which today hosts more than 7,000 datasets. We are also funding foundation models—some of which will be launched at the Impact Summit. These models will be open source and available to the ecosystem.

On top of compute, datasets and foundation models, we are building an orchestration layer—something like UPI for AI. If someone is building a healthcare AI application, instead of worrying about compute, datasets and model selection, they can plug into common infrastructure.

In many ways this is similar to the DPI that we have built. In practical terms, it exponentially expands the reach of DPI to include people who are otherwise unable to access online services. One can think of it as DPI raised to the power of AI.

DPI raised to the power of AI sounds really interesting. In terms of applications in different sectors that you talked about, are there going to be pilots or sandboxed initiatives, or will it be a rollout?

We are not looking at pilots or proof-of-concepts. Pilots often remain pilots and do not mature. If we have an AI application that can diagnose tuberculosis, cataract, diabetic retinopathy, or if we have AI tutors for students, or AI-based agricultural advisory services like Mahavistar in Maharashtra— then our goal will be deployment at population scale.

The objective of the IndiaAI Mission is rollout—institutional deployment—not limited pilots.

Looking ahead, what would define the success of IndiaAI Mission at the end of this decade—global competitiveness in foundational models, sovereign infrastructure depth, measurable economic contribution, or large-scale societal impact—and how will that success be measured?

Foundation models, sovereign infrastructure, datasets, compute—all of these are important. But they are means to an end. The real metric will be adoption of AI services by the majority of Indians—especially integrating those who are currently outside the digital economy.

Today, about 900 million people access the internet. Around 500 million do not access online services meaningfully. If a large proportion of these 500 million begin accessing services using AI tools—especially through voice-enabled interfaces—that will be real success.

If AI augments incomes, increases productivity, and brings efficiency gains to millions of citizens, that will be measurable economic and social impact. Everything else supports that goal. The ultimate measure of success is empowering millions of Indians through AI.

The interview was transcribed and edited with limited use of AI-based tool.