AI’s Real Test Is Diffusion, Not Disruption: Nandan Nilekani at AI Impact Summit 2026

At a time when artificial intelligence models are advancing at breakneck speed, the real challenge may not be building smarter systems but ensuring they reach everyone.

That was the central theme of a fireside chat between Dario Amodei, CEO of Anthropic, and Nandan Nilekani, co-founder and chairman of Infosys, at the AI Impact Summit 2026. The summit, underway since 14 February, featured the two technology leaders in conversation with Rahul Matthan, Partner, on how AI’s promise will ultimately hinge on diffusion, not just innovation.

Amodei drew a distinction between what AI systems are capable of today and how long it takes those capabilities to spread across economies.

“There is a duality between the fundamental capabilities of a technology and the time it takes for those capabilities to diffuse into the world,” he said.

He pointed to rapid advances in AI models that are increasingly strong at software engineering and are beginning to accelerate biomedical innovation. While the technology is progressing at what he described as a “very fast exponential,” he argued that the economic impact remains far below its potential.

Even if AI development were to freeze at today’s capability levels, Amodei suggested, “the economic impact could be much greater than it is.” The constraint, he said, lies in enterprise adoption. Companies face operational frictions, regulatory uncertainty, and institutional inertia that slow deployment. “This question of diffusion is tied with how everyone benefits from it,” he added.

India’s Playbook For Scale

If Amodei focused on technological acceleration, Nilekani focused on scale.

While praising the speed at which foundation models are evolving, Nilekani said diffusion is “a different ballgame.” India, he argued, has experience in taking digital infrastructure to population scale.

Citing the country’s digital public infrastructure, he noted that 1.4 billion Indians are enrolled in Aadhaar, and UPI processes over 200 billion transactions a month, forming what he described as the world’s largest cash transfer and financial inclusion system. The lesson, he said, is that diffusion is both “an art and a science.”

It requires institutions, policy-making, negotiations with incumbents and newcomers, and strategies for execution. “If all the investments in AI are going to deliver value to society, not just to individuals, we have to look at diffusion pathways to take this to everyone,” Nilekani said, reiterating his long-held view that India should aim to become “the use-case capital of the world.”

He also mentioned how the starting point must be the user.

If AI is to be treated as a general-purpose technology, he argued, the question is how it can tangibly improve lives: helping a billion people learn better, expanding access to healthcare, or increasing earnings for hundreds of millions of farmers.

“You have to start from there,” he said, and then work backwards to design the pathway. Technology is only one piece of the puzzle. Trust-building, institutional capacity, guardrails, stakeholder alignment, and data frameworks are equally crucial.

Diffusion, he stressed, is difficult. But India has already demonstrated the ability to operate at population scale across sectors. AI may introduce new complexities, especially around data governance, but multiple pathways can be built.

Nilekani warned that the global AI race is unfolding in two directions,  “a race to the top” and a “race to the bottom.” Those committed to ensuring AI serves humanity must double down on responsible diffusion. Otherwise, he cautioned, a backlash is inevitable.

Risks, Democracy And The Global South

Amodei also framed AI diffusion as particularly consequential for the Global South. In many developing economies, he argued, AI’s benefits , from productivity gains to service delivery improvements, could be disproportionately large.

But those benefits come with risks.

AI is a technology with “big risks and big benefits,” he said, cautioning that issues such as safety, predictability, and control cannot be ignored. As the world’s largest democracy, India has a critical role in shaping how democratic societies govern AI, particularly in contrast to authoritarian regimes deploying the technology.

Another concern is economic displacement. While Amodei said the “signature of the technology” will likely be to significantly expand the global economic pie, he acknowledged the potential for disruption in the process.

During his recent visit to India, he said he had been thinking about how global AI companies could work more closely with Indian firms. India, he observed, offers a “clean distillation” of both AI’s benefits and its risks.