Loading...
Building the full-stack sovereign AI platform for India.
Sarvam AI: The Startup Building AI for Bharat
What if India’s AI future didn’t have to be built elsewhere?
That question sits at the heart of Sarvam AI, a Bengaluru-based artificial intelligence startup founded in August 2023 by Dr. Vivek Raghavan and Dr. Pratyush Kumar. In just a few years, Sarvam has gone from an ambitious Indian AI research company to a $1.5 billion-valued startup, positioning itself around a particularly Indian challenge: building AI that understands how India actually speaks, works and communicates.
Sarvam’s founding team brought together two very different but complementary backgrounds.
Dr. Vivek Raghavan brought experience from India’s digital public infrastructure ecosystem, including his association with the technology architecture behind Aadhaar.
Dr. Pratyush Kumar, meanwhile, came from the research side of artificial intelligence and Indian-language computing, including work associated with AI4Bharat at IIT Madras.
Together, they combined experience in national-scale technology infrastructure and natural-language AI research—a combination that would eventually shape Sarvam’s central idea.
India should not merely consume AI built elsewhere. It should build AI for its own languages, people and institutions.
Sarvam attracted significant attention almost immediately.
In December 2023, the company raised around $41 million across its seed and Series A rounds, backed by investors including Lightspeed, Peak XV Partners and Khosla Ventures.
Then came the much bigger leap.
In June 2026, Sarvam announced the first close of its Series B at $234 million, with the round targeting $300 million. The funding valued the company at $1.5 billion, turning Sarvam into one of India's prominent AI unicorns. HCLTech committed $150 million as the lead strategic investor, while Bessemer Venture Partners, Khosla Ventures and Peak XV Partners also participated.
The capital isn't simply about building another chatbot.
Sarvam says it is building a full-stack sovereign AI platform—covering models, inference infrastructure and applications—so AI can be developed, deployed and governed within India.
One of Sarvam’s biggest differentiators is its focus on Indian languages and voice.
Its early foundational model, Sarvam-1, was a 2-billion-parameter model designed specifically for Indian languages. The company later open-sourced its 30B and 105B reasoning models, trained from scratch in India using compute provided through the IndiaAI Mission.
But language models are only one piece of the puzzle.
With Bulbul, Sarvam is developing expressive text-to-speech technology for Indian languages, designed to handle accents, regional variation and the reality of code-mixed Indian speech.
Its speech technologies—including Shuka and Saaras—focus on converting real-world Indian speech into usable text, even when conversations contain multiple languages or noisy audio.
Sarvam is also working on document intelligence and vision, targeting the enormous volume of PDFs, scanned records and multilingual documents used by Indian institutions.
Then there are Sarvam Agents—voice-based AI systems that can interact with people through telephones, WhatsApp and applications to perform business workflows.
There is a fundamental difference between building AI for people who comfortably type English prompts and building AI for a country where voice remains one of the most natural interfaces to technology.
That makes Sarvam's strategy particularly interesting.
Instead of asking every user to learn how to interact with an AI system, the company is trying to make AI adapt to the way people already communicate.
Speak naturally. Ask in your language. Let the AI handle the complexity.
Sarvam claims its technology is already being deployed at significant scale.
According to the company, its platform has supported voice interactions involving 17 million farmers for India's Ministry of Agriculture and a nationwide insurance campaign reaching 45 million policyholders. Sarvam also reports more than 2 million voice conversations per day, 10 million+ API calls per day and 35 million+ pages digitized.
The Government of India has also selected Sarvam under the IndiaAI Mission to build an indigenous sovereign large language model, providing dedicated compute resources for the project.
Sarvam's story isn't simply about raising hundreds of millions of dollars.
It is about a different proposition:
Build AI in India, for India—and make it capable of reaching India at population scale.
From two founders in 2023 to a $1.5 billion valuation in 2026, Sarvam is betting that India's AI opportunity won't be defined only by bigger models.
It may be defined by better models for India's languages, voices, institutions and everyday realities.
And that could make the next chapter of India's AI story considerably more interesting.
