New Delhi: Around 30-50 per cent of Sarvam AI’s latest round of $300 million funding will go towards procuring new compute and GPU power, the Bengaluru-based AI start-up that just became a unicorn, said. In an interview with ThePrint, Sarvam spoke about their $300 million Series B funding, which led to a billion-dollar valuation last week, and their plans for growth in India’s sovereign AI space.
“One of our focus areas will be frontier model research and development, including setting up a San Francisco office to attract top talent,” the company said in an email.
On 15 June, after their Series B funding closed, Sarvam joined the elite league of AI unicorns in India—with HCLTech leading the investment round.
Founded in August 2023 by Vivek Raghavan and Pratyush Kumar, Sarvam is developing full-stack AI platforms for governments, enterprises, and developers. The AI company also makes large language models and multimodal AI systems with a focus on Indian languages and region-specific use cases.
Their main selling point is that, unlike OpenAI, Anthropic, or Mistral, Sarvam AI is uniquely designed keeping the Indian customer in mind. They currently have five products out, including Sarvam Samvaad, which is a voice-based conversational AI available in 11 languages, and Sarvam Studio, which is an AI content tool for dubbing and translating.
The company shared that currently, more than 90 per cent of Sarvam’s usage is for revenue-generating business and government enterprises, rather than for individual customers. They plan to scale up their business by focusing on government, defence and enterprise clients, with an AI capability that is “built and served” in India. The recent restrictions imposed by the United States government on certain Anthropic models has sparked India’s AI sovereignty push.
“That means models that understand our voices, read our documents, and serve intelligence at a cost every enterprise and government can afford,” said Pratyush Kumar in a press release after their last round of funding.
According to Sarvam, the new round of funding will largely be used for expanding their compute infrastructure and scaling up their business through talent acquisition. While they did not set a date for the San Francisco office yet, the company did say that they will predominantly continue to hire in India, except for certain exceptional individuals based in the US.
Also Read: India is mistaking data localisation for digital sovereignty. It must control the traffic system
What are the company’s focus areas?
Over the next few months, Sarvam AI plans to focus on and expand two of its main products—Sarvam Samvaad for conversational AI, and Sarvam Akshar, an AI document digitalisation tool. For the former, they plan to scale up offerings to a full-scale software-as-a-service model. For the latter, the company is actively working to increase digitisation capabilities, especially tailored for Indian government documents.
The company is also going to build new agentic and coding products, beyond their existing five products.
However, Sarvam isn’t just building consumer-ready products; the full-stack approach means they are focusing on everything from the compute and cloud infrastructure and data centres, to the programming interface or APIs, training LLMs, and the final customer-friendly products. This is the “sovereign” aspect of Sarvam’s work. They are also in the league of the handful of labs globally that can pre-train competitive agentic AI models from scratch, according to Bessemer Venture Partners, one of the investors in Sarvam’s latest round.
While the current round of funding is enough for present capability enhancement, Sarvam also plans to raise additional rounds in the future, given the capital-intensive nature of the industry. The company was also selected as one of 12 organisations for the IndiaAI Mission, and received financial and computing support up to Rs 246.74 crore.
But Sarvam is also acutely aware of what works for it and what doesn’t. Their business is structured for sustainability and is modelled closer to Anthropic than to OpenAI.
Sarvam is clear that their business model is not direct B2C, the way Anthropic is; instead, their target customers are governments, the defence sector, large developers, and enterprises such as HCLTech. The company already has existing partnerships with agencies such as Swiggy for its Voice AI offerings, Pixxel for building orbital data centres, and even Microsoft for developing a sovereign, “Indic” LLM for Azure.
In terms of government customers, the company has partnered with the Unique Identification Authority of India (UIDAI) for Aadhaar services enhancement, and the Governments of Odisha and Tamil Nadu to build AI compute infrastructure and data centres.
“The Indian AI ecosystem is nascent but presents significant opportunities. Bessemer frames the coming decade as India’s AI decade. Sovereign AI is no longer optional,” the company said.
(Edited by Insha Jalil Waziri)