At ET AI Conclave & Awards 2025, Vivek Raghavan of Sarvam AI warns against India’s digital colonialism

At the ET AI Conclave & Awards 2025 on February 26, 2026, in Bengaluru, Vivek Raghavan, Co-founder of Sarvam AI, outlined why India must do the “hard things” and train models at scale.

ET Special
In a candid fireside chat at the ET AI Conclave & Awards 2025 on February 26, 2026, in Bengaluru, Vivek Raghavan, Co-founder of Sarvam AI, outlined the case for sovereign AI models amid global giants eyeing Indian data. Speaking with Samidha Sharma, Editor, ETTech, Raghavan, who is considered the architect of India's digital public infrastructure from Aadhaar to ONDC, positioned India at a geopolitical crossroads, likening the AI race to nuclear power dynamics.

The conversation opened with the buzz from the recently concluded India AI Impact Summit held from February 16-20, 2026, in New Delhi, where Sarvam unveiled foundational models built by a lean team of around 40 young engineers averaging 23-24 years old “I think we wanted to show that some of these foundational things are not just things that happen in the West Coast of the United States or happen in China,” Raghavan said, crediting the IndiaAI Mission's call for homegrown large language models (LLMs). He stressed this as proof that “India can compete at this skill, and we can build things that are world class and state of the art from India,” marking just the beginning of the story.

Sharma noted mixed reactions to the Summit: optics without big money versus mainstreaming AI for students and non-tech audiences. Raghavan made a case for Sarvam's resolution to scale up.


Sovereignty starts with auditable models
Transitioning to India's vulnerability as a data goldmine for frontier players such as OpenAI and Anthropic, Raghavan countered the risk of India being relegated to a “digital colony.” Invoking his Aadhaar-scale experience, Raghavan drew parallels to China, where top open-source models now dominate despite initial shocks from ChatGPT's 2022 debut. “Many of the top open source models in the world are actually Chinese,” he said.

Yet India differs, he argued, framing AI as a nuclear imperative: “There are the nuclear powers and there are the non nuclear powers. The United States and China will be frontier AI powers. There are a few countries which are in the middle, and India is one of them.” Sovereignty starts with auditable data in from-scratch training: “a model where the data that was used to train that model is auditable by the sovereign,” to guard against poisoning, as shown in Anthropic research. This layered approach extends to scaling inference domestically, where India still lags.

Scaling models, securing capital
Raghavan acknowledged frontier models' excellence: “they're wonderful, and they're getting better every month.” But he insisted smaller models meet 95% of needs when “put them in harnesses to actually do useful things,” ideal for billions of daily inferences. Sarvam's experiment unlocked “all the main knobs to actually build even a frontier model,” hinging now on capital.
ADVERTISEMENT

On funding post the February headline, he remained measured, focusing monetisation on business-to-business (B2B) and enterprise deals, plus strategic government sectors. “We are working with enterprises, B to B to C kind of, or B to B kind of. And we are also working with governments,” he said, downplaying grant dependency: “The vast majority are not competitive.” Inference at scale remains India's next frontier, challenging free global APIs.

Roadmap for India's AI builders
In closing advice, Raghavan urged focus on applications and orchestration: “There's so much value to be had at the application and at the orchestration layer. If the app is controlled by Indian companies, we can build things which are very useful for people in India.” Yet he circled back to necessity: “We need to do the hard things. We need to be able to train the models. We need to be able to run the models at scale. Otherwise we will be a digital colony.”

Sharma wrapped up: “Thanks a lot for doing that for India,” encapsulating the conclave's call to action.

The ET AI Conclave & Awards 2025 has L&T Finance as the NBFC partner, Snowflake as the AI Data Cloud Centre, EY as the Evaluation Partner, and T-Hub as the Ecosystem Partner, and is driven by BYD, with Indri as the celebration partner and Zoho as the Technology Partner. Vahdam India, Vaaree, Natch, and Andamen are the event’s gifting partners.
ADVERTISEMENT
Download
The Economic Times Business News App
for the Latest News in Business, Sensex, Stock Market Updates & More.
READ MORE
ADVERTISEMENT

READ MORE:

LOGIN & CLAIM

50 TIMESPOINTS

More from our Partners

Loading next story
Business News › AI › AI Insights › At ET AI Conclave & Awards 2025, Vivek Raghavan of Sarvam AI warns against India’s digital colonialism
Text Size:AAA
Success
This article has been saved

*

+