Cognizant Technology Solutions is in talks to take a small stake in Sarvam AI as part of a funding round that would value the Indian artificial intelligence startup at $3.5 billion to $4 billion, people aware of the matter told The Economic Times in a report on 28 October. Sarvam is seeking about $300 million in fresh capital.
A deal would put a second Indian IT services major on Sarvam’s cap table. HCLTech led the startup’s Series B in June with $150 million for a 10.46 per cent stake, and both sides framed the cheque as a distribution deal as much as a financial one. Neither Sarvam nor Cognizant has confirmed the new discussions.
Read more on newzshort:
Nvidia Backs Simplismart In $14 Million Series B Round
Why An IT Services Firm Wants A Model Lab
Cognizant’s interest fits the strategy its chief executive, Ravi Kumar S, set out on the company’s April earnings call. Acquisitions and investments, he said, should move the Nasdaq-listed services firm from billing for effort toward selling platforms and outcomes, the ambition it calls becoming an AI builder.
The incentive is defensive as well. Agentic AI, which can write code, maintain applications and run back-office processes, threatens the billable-hour model that built India’s outsourcing industry. Buying into the model layer gives a services firm a seat inside the technology that may compress its own work.
Read more on newzshort:
Legora Raises $550 Million Series E At $8.8 Billion Valuation To Expand Legal AI
The Arithmetic Behind The Valuation
The numbers explain the urgency. Sarvam closed the first $234 million of its $300 million Series B in June at a $1.5 billion valuation, with HCLTech putting in $150 million for 10.46 per cent and Bessemer joining existing backers Khosla Ventures and Peak XV Partners. A $3.5 billion to $4 billion mark would be 2.3 to 2.7 times that valuation in about four months.
Because most of the round is already funded, the conversation is about the last $66 million or so and who anchors it. A full $300 million at $3.5 billion implies dilution of 8.6 per cent for existing holders, and 7.5 per cent at $4 billion. The size of Cognizant’s slice has not been reported.
What Sarvam Actually Runs
Founded in 2023 by Vivek Raghavan and Pratyush Kumar, both formerly of AI4Bharat at IIT Madras, Sarvam had raised $41 million across seed and Series A before June. It released open-source models at 30 billion and 105 billion parameters this year, and in October added Bulbul V4 Flash for speech generation and Sarvam Vision 2.1 for document processing.
The company’s own usage figures, disclosed with the June round, give the business a shape: about 10 million API calls a day, 2 million conversational interactions daily, half a million hours of audio transcribed each month and 35 million pages of records digitised. Its voice agents collected data from 17 million farmers for the agriculture ministry.
It also runs a renewal campaign for an insurer covering 45 million policyholders, and an agentic platform for a large fintech’s sales force of 350,000 people. The company says the new money will fund models for agentic work, coding and cybersecurity, plus more compute.
A Market With One Clear Leader
Tracxn data published on 28 October put Sarvam’s share of funding raised by Indian language AI startups this year at about 87 per cent. That concentration cuts both ways. It makes the company the obvious partner for a ministry or enterprise wanting an Indian model, and leaves little room for a second vendor to build scale.
The imbalance sits inside a larger one. OpenAI and Anthropic both describe India as their second-largest market, yet the country has produced few companies attempting frontier models, held back by computing costs and capital. Sarvam is one of a handful, which is why strategic money keeps arriving at strategic prices.
Risks And Open Questions
The obvious risk is that the talks do not close. The report rests on people aware of the matter, the stake is described as small, and no term sheet has been confirmed. The valuation also rests on a round still being assembled, not a closed transaction.
There is a strategic question too. If IT services firms own pieces of the model layer, they become both distributor and potential casualty of the same technology. Sarvam’s pitch is that Indian language models, local data handling and government work are niches global labs will not serve directly. That thesis needs enterprise contracts, not only cheques.
What To Watch
Watch for confirmation from either company, the final size of the round and the remaining investors. Watch also whether Sarvam ships the agentic and coding models it has promised, and whether HCLTech’s distribution push shows up in its revenue.
