Sarvam AI's Vivek Raghavan says Big Tech 'dumping' free models, urges govt as first buyer 

India should treat heavily subsidised foreign artificial intelligence services as “dumping” and the government should step in as an early customer for domestic alternatives, according to Sarvam AI co-founder Vivek Raghavan, who argued that sovereign AI efforts will struggle without large-scale local adoption. 

“In any other industry, spending billions or hundreds of billions of dollars for something and then offering it for free would be considered dumping,” Raghavan told Business Today in an interview. “There should be some policy response.”  

His comments come as India accelerates efforts to build homegrown foundation models under the government-backed IndiaAI Mission, amid concerns that global tech giants could dominate the market by offering powerful AI tools at little or no cost. 

Raghavan said the economics of frontier AI make it difficult for startups to compete directly with multinational firms that have already invested tens of billions of dollars. “All of these frontier models are available for free to people today. As a B2C thing, it is very difficult for a startup like us to compete with that,” he said.  

Instead, he urged policymakers to ensure domestic models are widely deployed, particularly in public services, to create a sustainable ecosystem. 

“At the very least, we need to figure out how Indian models are being used at large scale,” Raghavan said, warning that without usage, local systems will not improve fast enough to remain competitive.  

He added that governments at both the central and state levels should take the lead as anchor customers. 

“It is important especially in a technology like this for, in some ways, the government to be a buyer of first resort, I mean as opposed to last resort,” Raghavan said. “That is what is going to make companies like us able to pay the bills.”  

Sarvam AI recently unveiled large language models designed for Indian languages and use cases, positioning itself as a key player in the country’s push for “sovereign AI.” The startup has received support under the IndiaAI Mission, including subsidised access to computing resources. 

Raghavan said the company has spent about $25-30 million to build its models, a fraction of the budgets deployed by global competitors, highlighting both the efficiency and the constraints facing Indian AI developers. 

The broader stakes extend beyond commercial competition, he argued, framing AI capability as a strategic necessity. 

“Sovereignty is important in the long arc… otherwise we will be a digital colony,” Raghavan said.  

For India, the question is not whether domestic firms can match Silicon Valley immediately, but whether the country chooses to remain a consumer of foreign AI systems or become a producer of foundational technology, he added. 

Without decisive policy support and early adoption, Raghavan suggested, even technically strong local models could fail to gain traction in a market increasingly shaped by global scale economics.