“We’re not just a model builder,” Pratyush Kumar, co-founder of Sarvam, told Mint on the sidelines of a gathering in Bengaluru last week. “We have an application layer too, and we want to be focused on both (the model and the layer).”
The startup has faced criticism after it last month released its AI model—Sarvam-M. While some hailed the 24-billion-parameter model for its native support for 10 Indian languages, others criticised it for not being a foundational model, but being based on a foundational model made by a French AI startup, Mistral.
Kumar and his investors, however, do not believe that the criticism has been fair. “It’s hard to quantify how much of our efforts we are devoting to foundational models versus building applications—the models will be key because this will give us the foundational capability. The government funding is just incidental, but it will not define the way forward for Sarvam. Eventually, we’re building a full-stack, deep-tech firm out of India,” he said.
To be sure, a foundational model is an algorithm built by a company from scratch, using its own data and AI infrastructure. Doing so requires significantly high amounts of capital—US Big Tech firms are currently spending billions on them—but also gives companies the advantage of having proprietary technology without being reliant on anyone else.
Monetising tech
An application layer, meanwhile, is key to ensuring that the models can be implemented within businesses, thus earning ventures commercial revenue. While this is secondary, it is key to monetisation. Big Tech firms, such as Google, Microsoft, and OpenAI, are making what Kumar cited above as “full-stack” services—having their own models and applications to sell to enterprises.
It is this that Kumar highlighted as Sarvam’s key area of focus, and not its government backing for subsidised access to processing chips—which account for most of an AI firm’s operating expenditure. “The government’s Graphics Processing Unit (GPU) grant helps, but it is only an episode for us. This is not what will fully define our way forward. We’re continuing to build and innovate in various sectors of AI, which will create value by leveraging the support that the government has extended. We’re building applications for Indian clients, and not just models or applications to sell abroad,” Kumar said.
Among its peers, investors and industry experts believe that Sarvam has a clear headstart. While Olacabs founder Bhavish Aggarwal’s AI pivot, Krutrim, claimed to have become “India’s first AI unicorn” in January last year, the venture doesn’t have much to show yet.
“Sarvam always had an aspiration of building population-scale AI. The goal is to scale up AI innovation to the entire Indian population, not to have the largest foundation model,” said Harshjit Sethi, managing director at venture capital firm Peak XV.
Peak XV led Sarvam’s seed funding round, and also co-led the startup’s $41-million series-A funding round in December 2023. The venture has raised $53 million in total funding so far, though Kumar said that it will “definitely need a larger funding round” in the coming months.
Peak XV’s Sethi underlined that the criticisms of Sarvam may not be accurate. “Sarvam is already working with some of the largest Indian companies, and has already reached tens of millions of individuals who have used Sarvam’s AI applications and models. The goal is bearing fruit, and we’re seeing their impact already.”
Use of funds
As a result, both Kumar and Sethi said that the public expectation that Sarvam is only building a ‘sovereign’ AI model for public utilities and government applications is misplaced.
“The government grant comes from public taxpayers’ money, but the one thing to clarify is that this is not a grant, the Centre is making an investment in the company in exchange for a minority equity stake. As a shareholder, the return to the taxpayer will come from the appreciation of this equity stake, in addition to the strategic imperative that we, as India, should have our own AI models from scratch,” Sethi said.
Kashyap Kompella, author and AI consultant to large corporations, concurred. “The entire criticism that Sarvam has faced is that it has the government’s backing from public taxpayers’ money, but its mission statement is not narrowly focused on building only India’s sovereign AI. That comes across as unfair criticism because innovation cannot be restricted only to one area. Entities like Sarvam must therefore focus on gaps in the ecosystem that are not addressed by competitors such as Big Tech firms,” he said.
To be sure, FY25 was the first full fiscal year for Sarvam as a commercial entity earning early-stage revenue from clients. The company is yet to file its financials for the previous fiscal. For now, Sethi said the startup is likely to make more product announcements “in the coming weeks.”
“There are no specific verticals that we’re focusing on, and we’re looking at technology and services quite horizontally, just like Big Tech ventures. Eventually, we’ll work with partners in sectors to build domain-specific capabilities in AI applications and services,” Kumar added.
