/vnd/media/agency_attachments/bGjnvN2ncYDdhj74yP9p.png)
/vnd/media/media_files/2026/02/19/at-ai-impact-summit-nilekani-and-amodei-debate-ai-diffusion-in-india-2-2026-02-19-15-34-19.jpeg)
At the India AI Impact Summit 2026, Nandan Nilekani, Co-founder of Infosys, and Dario Amodei, Chief Executive Officer of Anthropic, highlighted what Amodei described as a “duality” shaping artificial intelligence: exponential gains in model capability on one hand, and slow, friction-filled diffusion across enterprises and societies on the other.
In a session moderated by Rahul Matthan, Partner of Trilegal, the conversation focused on the widening gap between what frontier systems can technically achieve and how effectively institutions absorb them. Drawing on India’s experience in scaling digital public infrastructure such as Aadhaar and UPI, the speakers examined how diffusion determines economic impact.
Also Read: At AI Impact Summit, Modi shifts debate from tech to civilisation
Capability Races Ahead of Adoption
Amodei said models such as Claude are entering what he described as an exponential phase of capability in software engineering and biomedicine, yet the economic gains remain constrained by enterprise inertia.
“Even if we froze in place what the technology was capable of today, I think the economic impact could be much greater than it is because it just takes time,” Amodei said. “There are just frictions to adopt things through the enterprise.”
He said this divergence between technical progress and institutional readiness defines the current phase of AI. While the Global South could benefit from catch-up growth, the uneven distribution of gains could trigger resistance similar to the backlash against blue-collar globalisation.
Without broad-based deployment, he said, AI risks economic displacement and social disruption.
Also Read: Macron highlights India’s digital model at AI Summit
India as a Diffusion Laboratory
Nilekani responded by shifting the focus from model capability to system-wide implementation. While acknowledging the pace of development in foundation models, he argued that diffusion requires institutional coordination and trust.
“The speed of evolution in foundation models is impressive, but diffusion is a different ball game,” Nilekani said. “India will demonstrate this because we have the experience of diffusion at the population scale. It is both an art and a science. It involves institutions, policy making, negotiations, dealing with incumbents, dealing with newcomers, the whole trust-building thing.”
Rather than competing only in building frontier models, Nilekani said India should aim to become the “use case capital of the world”, concentrating on applications in education, healthcare and agriculture. If AI produces limited social value, he warned, a white-collar backlash is possible.
A key development at the summit was the launch of the “100 Diffusion Pathways by 2030” initiative, a coalition comprising Anthropic, Google, the Gates Foundation, and UNDP. The initiative aims to create open playbooks for scaling AI systems, combining technical guardrails, data strategies and institutional buy-in.
Also Read: Google unveils USD 15 bn AI data centre and cable plan at AI summit
Growth Ambitions and Inclusion
Amodei said Anthropic is expanding its engagement in India, citing a partnership with Infosys and advances in Indic language support. The company’s latest Claude release includes improved performance across 10 Indic languages.
“The fraction of Claude usage for technical programming and software engineering is substantially higher here in India than it is in most other places in the world,” Amodei said. “The use of Claude and Claude Code has doubled in India in just four months.”
Looking ahead, Amodei suggested that AI could help India achieve annual growth rates of 20–25% by leveraging its large population and technical workforce. Nilekani struck a more cautious tone, emphasising delivery over projections.
“I do not know about 25%, if I get 10%, I will be happy,” Nilekani said. “The focus has to be on inclusion. This AI has to work for people, starting from the user and how we can improve their lives.”
The Fireside Chat was transcribed and edited with limited use of AI-based tools.