The Equation Series
Menu

Artificial Intelligence · Profile from AI Research Companion

Sarvam AI

The Bengaluru start-up, founded in July 2023 by Vivek Raghavan and Pratyush Kumar, that set out to build a "full-stack" generative AI platform for India's languages and became the government's first pick to build a sovereign Indian foundation model under the IndiaAI Mission (April 2025). Sarvam's path shows the sovereign-model dilemma in miniature. Its first widely promoted large model, Sarvam-M (May 2025), was a fine-tune of a French open model and drew public ridicule for low downloads; nine months later, at the India AI Impact Summit in New Delhi (February 2026), it released two mixture-of-experts models trained from scratch, Sarvam 30B and Sarvam 105B, on 4,096 government-subsidised Nvidia H100s. Investors then made it India's second AI unicorn: a $300m Series B at a $1.5bn valuation, led by IT-services group HCLTech with $150m for ~10.5% (first close $234m, June 2026), completed with $74m including Nvidia's first investment (August 2026, reported). Revenue remains small: ₹45.1 crore (~$5m) in the year to March 2026 (reported). Sarvam is a young company, so this profile uses a set of smaller but well-sourced moves rather than padding a short history.

At a glance

Founded
July 2023, Bengaluru, by Vivek Raghavan and Pratyush Kumar
Headquarters
Bengaluru, India
Ownership / structure
Private (Sarvam AI Pvt Ltd, verify entity name). Investors: Lightspeed Venture Partners, Peak XV Partners, Khosla Ventures (from 2023); HCLTech (~10.46%), Bessemer Venture Partners (2026); Nvidia, Activate, Glade Brook Capital, Gaja Capital (August 2026) (reported)
Leadership (2026)
Vivek Raghavan (co-founder; earlier chief product manager and biometric architect for Aadhaar at UIDAI, and a volunteer at the iSPIRT think-tank, reported); Pratyush Kumar (co-founder; earlier IBM Research and Microsoft Research, co-founder of the AI4Bharat initiative at IIT Madras)
Layer
Model lab (sovereign, Indian languages) with voice-agent, API and enterprise platforms
AI products
OpenHathi (Hindi model on Llama 2, December 2023); Sarvam 2B and voice-agent platform (August 2024, verify); Sarvam-1 (October 2024, verify); speech models (Saaras/Saarika ASR, Bulbul TTS) and translation; Sarvam-M (May 2025, built on Mistral Small); Sarvam 30B and Sarvam 105B (18 February 2026, open weights); Samvaad conversational-agent platform; Sarvam for Work (announced 2026)
Scale
Revenue ₹45.1 crore (~$5m) in FY2025–26 (reported); >2m conversational interactions a day and ~10m API calls a day (June 2026, company, reported); headcount not disclosed (verify)
Capital raised
~$41m Series A led by Lightspeed with Peak XV and Khosla (December 2023, reported, verify); valuation ~$196m in 2025 (reported); Series B announced as $300m at $1.5bn post, ~$308m in total as reported (first close $234m on 16 June 2026; $74m completion reported 6 August 2026)
Government support
IndiaAI Mission: selected April 2025 to build the first sovereign LLM; 4,096 H100 GPUs for six months via Yotta; compute bill ~₹246.71 crore with a ~₹98.68 crore subsidy (reported); MoUs with Tamil Nadu and Odisha (February 2026)
Signature moves (incl. failures)
Founding and the full-stack Indian-language bet (2023); OpenHathi built on Llama (2023); voice agents and small models (2024); the IndiaAI sovereign-model selection (2025); the Sarvam-M backlash (2025); Sarvam 30B and 105B at the AI Impact Summit (2026); the HCLTech-led Series B and Nvidia (2026)

7 strategic moves analysed

Each one gets a full strategy file in the companion: context, the decision, the financing, what happened, and the lesson.

  1. 2023

    Founding and the Full Stack Indic Bet

    Founding / funding round

  2. 2023

    OpenHathi and Building on Llama

    Model release

  3. 2024

    Voice Agents and Small Indic Models

    Launch / model release

  4. 2025

    IndiaAI Sovereign Model Selection

    Government contract / compute deal

  5. 2025

    Sarvam M Launch Backlash

    Model release / failure (reputational)

  6. 2026

    Sarvam 30B 105B at the AI Impact Summit

    Model release

  7. 2026

    Series B HCLTech and Nvidia

    Funding round