Skip to main content

Sarvam AI

Category: AI Infrastructure

India's full-stack sovereign AI platform building generative AI solutions optimized for India's linguistic diversity, with voice-first, multilingual capabilities supporting 22 Indian languages for population-scale applications. Sarvam AI was founded in 2023. The company is led by Pratyush Kumar. Based in Bengaluru, India. Team size: 100-150. Total funding raised: $391.3M. Latest round: Series B ($300-350M, Apr 2026). Key investors include ["Bessemer Venture Partners","Lightspeed Venture Partners","Peak XV Partners","Khosla Ventures","Nvidia","Amazon","Prosperity7 Ventures"].

Founded
2023
Headquarters
Bengaluru, India
Team size
100-150
Total funding
$391.3M

Value proposition

Enables organizations to deploy AI at population scale through voice-first, multilingual interfaces that work for India's diverse linguistic landscape—breaking access barriers for the next billion users who cannot read, write, or type in English, while maintaining data sovereignty and enterprise-grade security.

Products and solutions

["Sarvam 105B - Flagship large language model","Sarvam 30B - High-performance multilingual LLM","Sarvam 2B - Compact open-source Indic LLM","Saaras V3 - Speech-to-text for 22 Indian languages","Bulbul V3 - Text-to-speech across 11 Indian languages","Samvaad - Enterprise conversational AI agents platform","Sarvam Agents - Voice-enabled multilingual business agents","Sarvam Vision - Vision-language model for document digitization","Sarvam Translate - Translation model for 22 Indian languages","Sarvam Dub - Multilingual dubbing service","Sarvam Kaze - AI-powered wearable glasses","Sarvam Studio - Application development platform"]

Unique value

Building India's sovereign AI stack from the ground up with voice-first architecture supporting 22 Indian languages; combines frontier-class models with population-scale deployment capabilities; open-source approach with Apache 2.0 licensing; government-backed initiative under IndiaAI Mission with access to 4,096 NVIDIA H100 GPUs; superior token efficiency (fertility rates 1.4-2.1) making models 2-4x more efficient for Indian languages than multilingual alternatives.

Target customer

Indian enterprises requiring multilingual AI solutions, central and state government agencies, public sector organizations, banks and financial institutions, healthcare providers, e-commerce platforms, developers building India-focused applications, and startups in the AI ecosystem.

Industries served

["Financial Services & Banking","Government & Public Sector","Healthcare","E-commerce & Retail","Enterprise Software","Telecommunications","Legal Services","Education Technology","Digital Public Infrastructure"]

Technology advantage

Full-stack integration from foundational models to enterprise applications; curated training corpus of 4 trillion tokens with 2 trillion high-quality Indic tokens; hybrid architecture combining speech, text, and vision capabilities; partnerships with UIDAI for Aadhaar integration and state governments for sovereign compute infrastructure (50MW AI-optimized capacity hubs); cost-effective voice agents starting at ₹1/minute; models designed for reasoning, instruction-following, and real-world Indian use cases rather than just English benchmarks.

How they differentiate

Sovereign AI stack with voice-first architecture for 22 Indian languages; government-backed IndiaAI Mission partnership; open-source model releases (Apache 2.0); superior token efficiency (fertility rates 1.4-2.1) for Indian languages; full-stack integration from foundational models to enterprise applications; curated 4 trillion token training corpus; population-scale deployment capabilities through UIDAI Aadhaar partnership

Main competitors

["Krutrim AI","Gnani.ai","CoRover.ai"]

Key partnerships

["Government of India (IndiaAI Mission, MeitY) - Selected to build sovereign foundational model","UIDAI (Aadhaar) - Voice-based AI services for 1.4 billion citizens","State Governments of Odisha & Tamil Nadu - 50MW AI-optimized compute infrastructure","Microsoft Azure - Hosting Indic voice LLMs on Azure platform","NVIDIA - Hardware support and GPU infrastructure (4,096 H100 GPUs)","AI Alliance (Meta, IBM) - Industry consortium participation","HMD Global - Hardware integration for Bulbul voice AI and Kaze glasses","IIT Madras - Research collaboration and Digital Sangam AI Research Park","Yotta Data Services - GPU infrastructure hosting"]

Notable customers

["UIDAI (Aadhaar - 1.4B citizens)","SBI Life Insurance","HCLTech","Razorpay","Government of Odisha","Government of Tamil Nadu"]

Major milestones

["Selected by Government of India under IndiaAI Mission to build sovereign foundational model (Apr 2025)","Partnership with UIDAI for AI-driven voice-based Aadhaar services (Mar 2025)","Open-sourced Sarvam-30B and Sarvam-105B models under Apache 2.0 license (Mar 2026)","Received 4,096 NVIDIA H100 GPUs and ~₹99 crore GPU subsidies","Sarvam Vision OCR achieved 84.3% accuracy, outperforming ChatGPT and Gemini on Indian language documents"]

Growth metrics

114 employees (Aug 2025) representing 226% YoY growth; selected by Government of India for IndiaAI Mission; partnership with UIDAI for 1.4B citizens; Sarvam Vision OCR outperformed ChatGPT & Gemini on Indian language documents

Market positioning

India's sovereign AI infrastructure leader with government backing, positioned as the national champion for Indic language AI competing with global giants like OpenAI and Google while differentiating through India-first approach and multilingual capabilities

Geographic focus

India (primary), expanding to global markets with Indic diaspora focus

Patents and IP

No registered patents publicly disclosed as of latest update; primary IP strategy focuses on open-source model releases under Apache 2.0 license and proprietary enterprise platform capabilities.

About Pratyush Kumar

Ph.D. in Computer Engineering from ETH Zurich (2009-2014); B.Tech in Electrical and Electronics Engineering from IIT Bombay (2005-2009); Former Researcher at Microsoft Research and IBM Research (2015-2022); Adjunct Faculty at IIT Madras (2017-Present); Co-founder of AI4Bharat (2020-2023) and OneFourth Labs (PadhAI); Published 89+ research papers at top-tier AI conferences; Over 15 years of research and development experience in AI/ML systems, Natural Language Processing, and multilingual AI for Indian languages.

Official website: