For decades, India has been one of the largest markets for artificial intelligence technology that originated from Silicon Valley, ran through US servers, and was regulated out of Washington, DC. However, this all changed on June 15, 2026, when the Indian startup AI company named Sarvam managed to raise $234 million in funding while earning a valuation of $1.5 billion, becoming an Indian AI unicorn.
Sarvam was jointly founded in 2023 by Vivek Raghavan and Pratyush Kumar, who have declared their mission to be promoting the accessibility of AI technologies in India, with the emphasis on creating models that are tailored specifically to India’s linguistic and cultural environment—with all products being developed, operated, and governed completely in India.
Both founders were working at the time on the AI4Bharat project, which was pioneering research on Indian language AI conducted at the IIT Madras. They have never intended to create a chatbot or a consumer application. Their goal was to solve an issue that none of the American companies could address properly—making AI technologies work in India’s unique linguistic environment.
Sarvam AI presents itself as a full-stack AI company. Unlike numerous other startups that concentrate exclusively on applications, Sarvam creates AI models, infrastructure, and enterprise solutions simultaneously.
In the middle of June, more specifically on June 15, 2026, the AI solutions developer called Sarvam reported that they have completed fundraising in the first installment of their Series B amounting to $234 million and have a target of $300 million total with the company valued at $1.5 billion post-money. Below is the list of investors involved in this deal:
HCLTech, lead strategic investor, is investing $150 million and acquiring a 10.46% share of the company.
Bessemer Venture Partners – the new investor joining the series;
Khosla Ventures & Peak XV Partners – existing investors returning for an investment.
The idea is to integrate AI solutions of the young company with HCLTech's enterprise connections, engineering team, and software infrastructure to create products for enterprises and governments using AI technologies.
This is much more than just investment. This is an alliance of the rising AI power of India and one of its strongest IT companies.
This is the core question behind everything that Sarvam stands for. This question was answered definitively in June 2026.
With the United States government directing Anthropic to limit the foreign access to its leading-edge models, firms outside of America received an unambiguous lesson that your AI stack belongs to you until the rules change. Firms in India using American frontier models understood, all of a sudden, the risk that they had been carrying around all along.
The structural logic behind India’s AI opportunity is similarly strong. India boasts 22 official languages, none of which gets adequate support from existing multilingual AI frameworks – something that American hyperscale companies cannot easily fix from abroad. The sheer size of its 1.4 billion population, where tens of millions of people access governmental services using digital infrastructure, provides enough of an addressable market for building indigenous AI to justify the investment.
To put it simply, if you need AI that really works for a farmer in Tamil Nadu, a customer of a bank in Gujarat, or a government office in Odisha, you can’t just translate ChatGPT. You have to start from scratch.
Sarvam's technical evolution has certainly faced some criticism. As early as May 2025, Sarvam introduced its first model, which was created using an already existing French model by optimizing it. It was criticized severely by some critics and even some important investors who said that it was not at all indigenous.
This criticism was well taken by Sarvam. In February 2026, Sarvam 30B and Sarvam 105B were trained from scratch using pre-training, supervised fine-tuning, and reinforcement learning processes, all done in India through the IndiaAI Mission.
Sarvam has also introduced Sarvam Vision, a model optimized for handwriting detection and recognition in Indian languages, which is being used for digitizing more than 35 million pages.
The government was involved as well. The Ministry of Electronics and Information Technology, India, identified Sarvam as one of the twelve institutions under the IndiaAI Mission to develop indigenous foundational AI models. It gave them 4,086 Nvidia H100 GPUs for six months, which is the largest computing subsidy ever provided by the mission in its first phase.
This is not just a research lab with stellar benchmarks. This is already live in places that matter.
The company’s conversational AI platform is already handling more than 2 million conversations every single day, while the inference platform handles around 10 million API calls every day.
The company’s speech models transcribe more than 500,000 hours of audio every month.
The true impact is also pretty staggering. The multilingual voice agents of Sarvam have helped collect data from 17 million farmers on behalf of India’s Ministry of Agriculture and Farmers Welfare. A countrywide voice-based campaign by a leading insurer has helped 45 million policyholders' policy renewals. One big fintech company is leveraging Sarvam’s Agentic AI Platform for a sales team of more than 350,000.
This is not just piloting. This is a national-scale deployment.
The new funding will allow for training of Sarvam’s next generation of frontier models that will concentrate on applications in agentic, coding, and cybersecurity areas while increasing computing capabilities to scale its operations in several verticals.
Sarvam is also planning a complete Series B funding round for $300 million in the coming months.
Co-founder of Sarvam, Vivek Raghavan, described the ambitions of the project in the following way during its announcement: "Our objective is to enable wide-scale diffusion of this technology in India, providing immense value across sectors for individuals, small businesses, corporations, and states & central governments."
