“India cannot afford to be a consumer of AI built elsewhere — it must build its own.” — The strategic logic behind Sarvam AI’s sovereign model
On 15 June 2026, Bengaluru-based Sarvam AI became India’s newest unicorn after raising $234 million in the first close of its Series B funding round at a post-money valuation of $1.5 billion. The round was led by HCLTech with a $150 million strategic investment — the most significant capital infusion yet for a domestically-built Indian AI company.
Founded in August 2023 by Dr Vivek Raghavan and Dr Pratyush Kumar, Sarvam AI is a full-stack sovereign AI startup focused on Indian-language AI. In April 2025, it became the first company selected — from 67 applicants — to build India’s sovereign Large Language Model (LLM) under the IndiaAI Mission, backed by a government outlay of ₹10,371 crore.
🌍 What Is Sovereign AI and Why Does It Matter?
Sovereign AI refers to artificial intelligence systems developed using domestic data, domestic computing infrastructure, and models trained within the country’s borders — as opposed to relying on foreign cloud providers, foreign-trained models, or data pipelines governed by overseas jurisdictions.
For India, with 22 officially recognised languages, over 800 million citizens more comfortable in regional languages than English, and a world-class Digital Public Infrastructure stack (Aadhaar, UPI, DigiLocker), the case for sovereign AI is both strategic and practical. Models trained primarily on English-language internet data systematically underperform on Indian-language tasks, regional accents, code-mixed speech (such as Hinglish), and culturally specific contexts. Sarvam AI was explicitly built to close this gap.
Think of Sovereign AI like having your own UPI instead of depending on PayPal. Just as India built its own payments infrastructure that works for Indian users in Indian languages, Sarvam AI is building AI that works for India — trained on Indian data, in Indian languages, running on Indian servers. You own the model; you own the data; you own the future.
India’s Digital Public Infrastructure (Aadhaar, UPI, DigiLocker) gave India financial and digital sovereignty. Sarvam AI and the IndiaAI Mission are attempting the same for artificial intelligence — the next layer of critical national infrastructure. The question is: can India replicate the UPI success story in AI at global scale?
👤 Founders: The Architects Behind Sarvam AI
Dr Pratyush Kumar (CEO) is an alumnus of IIT Bombay (BTech, Electrical Engineering) and completed his PhD in computer engineering from ETH Zurich in 2014. He worked as a research scientist at IBM Research and Microsoft Research before joining IIT Madras as faculty (2018–2021). He co-founded AI4Bharat — an open-source Indian-language AI initiative at IIT Madras — and One Fourth Labs, an AI education venture.
Dr Vivek Raghavan (Co-founder) graduated from IIT Delhi and earned his PhD in Electrical and Computer Engineering from Carnegie Mellon University. He served as Chief Product Manager and biometric architect at UIDAI — where he helped design and deploy the Aadhaar platform at population scale. His experience with India’s foundational digital identity infrastructure directly informs Sarvam’s ambition to build population-scale AI systems. Both founders were previously associated with AI4Bharat at IIT Madras before launching Sarvam in August 2023.
| Detail | Dr Pratyush Kumar (CEO) | Dr Vivek Raghavan (Co-founder) |
|---|---|---|
| Undergrad | IIT Bombay (BTech, EE) | IIT Delhi |
| PhD | ETH Zurich (2014) | Carnegie Mellon University |
| Notable Past Role | IBM Research, Microsoft Research, IIT Madras faculty | Biometric Architect, UIDAI (Aadhaar) |
| Pre-Sarvam Venture | AI4Bharat, One Fourth Labs | AI4Bharat (IIT Madras) |
💰 Series B Funding: Structure and Strategic Significance
The Series B round is structured in multiple closes, with the first close of $234 million announced on 15 June 2026. HCLTech — India’s third-largest IT company by revenue — invested $150 million and acquired a 10.46% stake in Axonwise Ltd, the operating entity behind Sarvam AI. This is a strategic, not passive, investment: HCLTech will integrate Sarvam’s sovereign AI capabilities across its enterprise client base in India and globally.
Bessemer Venture Partners participated as a new institutional investor. Existing backers Khosla Ventures and Peak XV Partners also participated. The total targeted Series B size is $300 million, with $66 million still to be raised. The capital will primarily be deployed to train frontier AI models in three domains: agentic AI, coding, and cybersecurity.
✨ Products & Model Architecture
Sarvam AI has built a vertically integrated product stack spanning foundational models, speech systems, developer APIs, and enterprise applications:
- 30B Parameter Model: Supports real-time conversations with a 32,000-token context window; trained from scratch in India for Indian languages
- 105B Parameter Model: Handles complex, longer-context tasks with a 128,000-token window
- Sarvam Vision: A 3-billion parameter vision-language model for multilingual document intelligence — interprets charts, nested tables, and scanned documents across Indian scripts
- Bulbul V3: Text-to-speech engine covering 12+ Indian languages with production-grade voice output across regional accents
- Sarvam Arya: Conversational AI platform for enterprise deployments
All models are accessible via Sarvam’s API platform priced in Indian Rupees — a deliberate advantage over foreign providers whose USD-denominated APIs attract 18% GST as a reverse charge for Indian businesses.
Three Sovereign LLM Variants under IndiaAI Mission: Sarvam-Large (advanced reasoning), Sarvam-Small (real-time interactive), Sarvam-Edge (on-device/compact). Developed in collaboration with AI4Bharat at IIT Madras.
⚖️ IndiaAI Mission: The Government Mandate
In April 2025, the Government of India selected Sarvam AI as the first company — chosen from a pool of 67 applicants — to build India’s sovereign Large Language Model under the IndiaAI Mission. The mission was approved in March 2025 with a budgetary outlay of ₹10,371 crore (~$1.2 billion), aimed at building domestic AI infrastructure, reducing reliance on foreign platforms, and training AI talent at scale.
Under the contract, Sarvam received access to 4,096 NVIDIA H100 GPUs for a six-month initial development phase, along with government subsidies worth approximately ₹99 crore for compute infrastructure. The sovereign LLM is being developed in collaboration with AI4Bharat at IIT Madras and will be built, deployed, and optimised entirely within India’s borders.
Don’t confuse dates: The IndiaAI Mission was approved in March 2025; Sarvam was selected under it in April 2025. The open-source 30B/105B models were launched in February 2026. The unicorn status was achieved on 15 June 2026. These are four distinct events — all frequently tested individually in MCQs.
📌 Commercial Deployments: Scale & Sector Coverage
Sarvam AI has moved well beyond research to active large-scale deployments:
- 17 million farmers: Multilingual voice agents for data collection deployed for the Ministry of Agriculture and Farmers Welfare — voice-first AI in rural, low-connectivity environments
- 45 million policyholders: Insurance policy renewals for a leading Indian insurer
- Sarvam Startup Programme: Offers ₹10 crore in free API credits to startups to build an ecosystem of developers and enterprises
Key sectors served: fintech, agritech, healthtech, and government services — all areas where regional language support and voice-first interfaces are essential for reaching Tier-2 and Tier-3 users.
🌍 India’s Unicorn Ecosystem: Context
A unicorn is defined as a privately held startup valued at $1 billion or more — a term coined by venture capitalist Aileen Lee in 2013. As of mid-2026, India has approximately 130 unicorns, making it the 3rd largest unicorn ecosystem globally, behind the United States and China. Bengaluru alone accounts for 53+ unicorns, confirming its position as India’s dominant deep-tech hub.
In 2026, Sarvam was the 5th new unicorn, preceded by Skyroot (space technology, May 2026). Sarvam joins a deep-tech cohort alongside Neysa (AI cloud infrastructure, $1.4 billion valuation, raised $600 million led by Blackstone in February 2026) and Skyroot Aerospace. Globally, AI businesses accounted for 12 of 47 new unicorns in 2026 (as of mid-year) — approximately 25.5% of the total.
Don’t confuse Neysa and Sarvam: Both are Indian AI unicorns in 2025–26, but they are different companies. Neysa focuses on AI cloud infrastructure (valued at $1.4B, raised $600M led by Blackstone in Feb 2026). Sarvam focuses on sovereign LLMs and Indian-language AI (valued at $1.5B, raised $234M led by HCLTech in June 2026).
Click to flip • Master key facts
For GDPI, Essay Writing & Critical Analysis
5 questions • Instant feedback
HCLTech led the Sarvam AI Series B with a $150 million investment, acquiring a 10.46% stake in Axonwise Ltd — the operating entity behind Sarvam AI.
Sarvam AI was selected first from 67 applicants in April 2025 to build India’s sovereign LLM under the IndiaAI Mission, which was approved in March 2025 with ₹10,371 crore outlay.
Dr Vivek Raghavan earned his PhD from Carnegie Mellon University and previously served as biometric architect at UIDAI, where he helped design the Aadhaar platform at population scale.
India is the 3rd largest unicorn ecosystem globally with approximately 130 unicorns as of mid-2026, behind the US and China. Bengaluru accounts for 53+ unicorns.
Sarvam AI launched its open-source 30B and 105B parameter models in February 2026 — both trained from scratch in India for Indian languages and local use cases.
