India needs at least a few truly powerful national frontier models if it is to achieve genuine technological self-reliance | Photo Credit: LindaMarieB

Frontier AI models have crossed a threshold. They are instruments of national power, economic dominance, and strategic advantage. As their capabilities grow, governments seem to be becoming less willing to leave decisions about access entirely to laboratories and markets. The perimeter of export control is expanding — from advanced chips and manufacturing equipment towards access to the models themselves.

The power these models carry is extraordinary. They can accelerate pandemic preparedness, democratise world-class education, raise farmer incomes through precise advisory systems, and enable genuine cross-border collaboration that lifts prosperity for everyone. Few technologies have offered comparable potential for broadly shared human progress.

Yet greater capability has not produced a commensurate commitment to shared access. Instead, it has intensified the race for strategic advantage. Governments and companies are erecting walls around capabilities that could otherwise accelerate development across the world.

So, is the real danger that AI will become too powerful… or is it that its power will become too concentrated in too few hands?

Diffusing intelligence?

As we navigate the road ahead, we must stop wasting precious time debating diffusion versus development. Democratisation of intelligence inside a nation’s own borders is essential for economic dynamism and societal benefit.

But when the core models that matter most are controlled by a handful of labs and the governments that regulate them, it forces an uncomfortable question: are countries diffusing intelligence they truly control — or merely distributing imported capability that can be withdrawn when a crisis or geopolitical calculation demands it?

This is not an either-or decision. Sovereign production of intelligence is non-negotiable. It gives us the ability to turn a nation’s own data, institutions, talent, and strategic priorities into world-class frontier capability that no foreign power can switch off.

Some also argue that India should concentrate on smaller, vertical and domain-specific models rather than compete at the frontier. The first part of that argument is right. The conclusion is not.

Smaller models are already demonstrating substantial value in agentic systems. They can be developed faster, operated more cheaply and aligned more precisely with particular languages, industries and public priorities. India should move aggressively to build and deploy them in healthcare, agriculture, education, financial services, law, manufacturing and public administration. But India also needs at least a few truly powerful national frontier models if it is to achieve genuine technological self-reliance.

India is not starting from zero. Organisations such as Sarvam AI, BharatGen, Fractal AI and others have demonstrated serious technical ambition and an understanding of India’s linguistic and developmental requirements. And I am sure the country can mobilise the necessary financing. The real task is prioritisation and ruthless channelisation. Money must move at the speed of the technology, not the speed of normal bureaucratic processes. And, we need the right talent.

Leading architects

Indians are among the leading model architects and researchers powering frontier laboratories around the world. One of the highest-leverage moves India can make is to create compelling reasons for some of this talent to return — and to build in India and for India. Embed them within the best Indian research institutions. Give them the mandate, the compute, the data access, and the decision rights to deliver a national frontier model on an accelerated timeline.

Last but not the least, government must create the bridge to markets. Even the best model is worthless if it never reaches real markets and real users at scale. This is where government has its most important and least appreciated role: market connection.

We are watching a predictable but unfortunate pattern. With great power has come not great responsibility, but the race to capture maximum value and then close the barn door so others cannot catch up.

The result is a rapidly hardening divided global order: a small club of AI haves and a large periphery of have-nots. This fragmentation will produce mistrust, duplicated effort, lower global growth, and heightened geopolitical risk.

As G7 leaders meet this week, the larger question is what kind of global order they are creating. The technology now exists to generate unprecedented prosperity and problem-solving capacity for humanity. It can also be used to lock in permanent advantage and permanent division. Thankfully the choice remains available.

The writer is Distinguished Fellow NITI Aayog and former President, Nasscom

Published on June 16, 2026