“This Is Year One For Us”: Sarvam’s Pratyush Kumar On India’s AI Opportunity
Last Updated: 2026-09-26 12:16:03
India’s AI story may have already produced unicorns, but Sarvam co-founder Pratyush Kumar believes the country is still at the very beginning of its journey in artificial intelligence.
Speaking at the Moneycontrol Startup Conclave 2026 , Sarvam co-founder Pratyush Kumar described the current phase of India’s AI development as “year one”, with Indian companies only now beginning to build models, computing infrastructure and products at meaningful scale.
Latest and Popular Mobiles
Kumar was speaking alongside Mukund Jha, Founder of Emergent, during a panel titled “India’s AI Unicorns: Models, Moats & the Race to Scale”, moderated by Nalin Mehta, Managing Editor, Moneycontrol.
The discussion covered India’s ambitions in AI, the development of indigenous models, access to GPUs, AI agents, enterprise adoption, open-source models and the security risks associated with increasingly capable AI systems.
Related Articles
India’s AI Race Is Only Beginning
For Kumar, the most important shift is that India has moved from being primarily a user of AI technology to beginning to build its own AI capabilities.
“This is year one for us,” he said during the discussion.
Kumar pointed to Sarvam’s own work as an example, saying the company has launched its first set of models. He also noted that only a handful of companies outside the US and Chinese ecosystems have trained models of 100 billion parameters or more from scratch.
That does not mean Kumar believes India has already closed the gap with the leading AI ecosystems. Instead, he sees the country’s scale, talent and willingness to adopt technology as potential advantages as the industry develops.
He compared the opportunity with India’s digital payments transformation. Kumar noted that India was not among the world’s top 10 digital payments markets in 2016, but has since become a major centre of digital payments.
He believes AI could move even faster.
“Five years down the line, India would be the largest user by far of this technology,” Kumar said.
Why Cost Could Matter As Much As Capability
For India, Kumar said, simply having access to powerful AI models will not be enough. The country also needs to be able to use them at scale and at a cost that works for Indian users and businesses.
That is one reason Sarvam has been investing beyond model development.
Kumar said the company is operating its own GPU cluster and has been building an inference stack that can host both Sarvam’s models and other open models. The goal is to provide AI inference in India at costs comparable with global providers.
The company is also developing products across areas including documents, voice, video, coding and cybersecurity.
Kumar said Sarvam increasingly sees itself as a full-stack AI provider, with the aim of giving Indian enterprises much of what they need to adopt AI at scale.
AI Needs To Solve Indian Problems
The Sarvam co-founder also argued that India’s AI opportunity extends well beyond consumer-facing chatbots.
He pointed to agriculture, healthcare and education as areas where AI could provide access to intelligence at population scale.
Kumar referred to work involving farmers and the digitisation of forms as examples of how AI models can be adapted for large-scale use cases in India.
He also argued that Indian businesses are demanding more from AI while remaining highly focused on cost.
According to Kumar, most of Sarvam’s revenue comes from the private sector, and the company is focused on building products that Indian enterprises are willing to pay for.
“AI is an ROI positive component for every enterprise,” Kumar said, adding that the challenge is taking AI and embedding it into businesses so that it actually works for them.
Emergent Is Betting On AI Agents
While Sarvam is building across models and infrastructure, Emergent is increasingly focused on what happens when AI becomes capable of performing entire business functions.
Mukund Jha said Emergent began as a platform focused on automating software creation but is now expanding towards business automation.
He said companies are already using its agents for functions including finance, marketing and growth.
Jha’s longer-term vision is for businesses to have persistent AI agents operating in the background, with agents from different departments communicating with one another and handling large parts of business operations.
He said more than 30,000 businesses are currently using one of Emergent’s persistent agent products daily.
Jha also said the rapid improvement in open-source models has become important for Emergent, particularly as reasoning and coding capabilities improve.
The Faster AI Gets, The Bigger The Security Question
The panel also highlighted the other side of India’s AI opportunity: security.
Jha said AI models are becoming increasingly capable of identifying and exploiting vulnerabilities, making investment in defence and security more important.
He argued that the risks could increase as future generations of models become smarter.
Kumar similarly distinguished between AI’s technological capabilities and the risks created when those capabilities are used deliberately for harmful purposes.
He said India needs its own defensive capabilities, particularly because so much of the country’s infrastructure is now digital.
At the same time, he argued against India slowing down its AI development simply because of these risks.
“From the India vantage point, we should accelerate,” Kumar said.
His argument was that India needs to pursue AI development while building the security and defence capabilities required to protect the systems that increasingly depend on the technology.
The Next Five Years Could Define India’s AI Journey
The discussion at Moneycontrol Startup Conclave 2026 highlighted two different but connected parts of India’s AI race.
Sarvam is working on models, GPUs, inference infrastructure and enterprise products, while Emergent is betting on AI agents that can increasingly automate business functions.
For Kumar, however, the larger opportunity is India’s ability to deploy AI at enormous scale.
The country still has a gap to close with the world’s leading AI ecosystems, but the foundations of an Indian AI industry are now being built.
That is why Kumar’s “year one” description is significant. His argument is not that India has already arrived in the global AI race, but that the country is only now entering the phase where it can begin building and scaling its own capabilities.
The question for the coming years, therefore, is how quickly India can turn that emerging capability into widespread adoption while also building the infrastructure and security systems needed to support it.