Sarvam builds Indian-language AI models and enterprise tools for voice, text, and document workflows. The Sarvam AI funding round has pushed the Bengaluru startup into unicorn territory after it raised $234 Mn in the first close of a $300 Mn Series B at a $1.5 Bn post-money valuation, with HCLTech leading the round and committing $150 Mn. That matters because Indian enterprises don’t just want a flashy chatbot anymore — they want AI that can handle local languages, plug into real systems, and run in production. Founded in 2023 by Pratyush Kumar and Vivek Raghavan, Sarvam is now using fresh capital to bet harder on agentic AI, coding, cybersecurity, and bigger compute capacity.
What does Sarvam AI actually sell?
Sarvam isn’t one app. It’s a full-stack AI company that sells models and APIs. It also sells enterprise products tuned for Indian languages. Its flagship conversational platform, Sarvam Samvaad, lets businesses deploy agents across voice calls, WhatsApp, and the web. Those agents connect to enterprise systems so they can pull customer data, trigger actions, and return conversation analytics instead of just answering prompts.
Under that application layer sits a broader developer stack. Sarvam exposes speech-to-text, text-to-speech, translation, chat completion, and document digitisation APIs. Named models include Saaras v3 for ASR and Bulbul v3 for TTS. It also offers Mayura and Sarvam-Translate for multilingual translation, Sarvam-30B and Sarvam-105B for chat, and Sarvam Vision for OCR and structured document extraction in 23 languages including English.
The customer workflow is pretty practical. A team can get an API key in minutes and test a pilot fast. Then it can move to managed cloud or private VPC-style deployment depending on security needs. Samvaad also keeps context across channels, so a user who starts on a phone call and comes back on WhatsApp doesn’t feel like they’re talking to a fresh bot every time.
That’s the sales pitch. Sarvam is trying to replace the usual pile of separate speech vendors and translation tools. It also wants to replace bot orchestration layers and analytics add-ons with one platform built for code-mixed Indian usage. Sarvam’s own 30B model already powers Samvaad.
Who built Sarvam AI and why now?
Founding story
Sarvam was founded in August 2023 by Dr. Vivek Raghavan and Dr. Pratyush Kumar. Their thesis was simple and pretty aggressive: India needed its own AI stack instead of depending entirely on foreign model providers, especially for multilingual use cases and regulated industries where data control matters.
Why the founders fit this market
Raghavan brings infrastructure instincts, not just AI enthusiasm. He’s closely associated with India’s digital public infrastructure work — from Aadhaar to ONDC — and earlier served as managing director of Synopsys India after returning to the country in 2007. Sarvam isn’t selling consumer novelty. It’s trying to build systems that can survive population-scale and enterprise-grade workloads.
Kumar brings the research depth. He studied electrical engineering at IIT Bombay and earned a PhD from ETH Zurich. He also worked at IBM and Microsoft Research, and served as an adjunct faculty member at IIT Madras. Sarvam says he led India’s open-source AI efforts across Indian languages.
Track record and early signals
The product timeline shows a team moving fast, not drifting. Sarvam started with OpenHathi-v1 in December 2023 — a Hindi model fine-tuned on Meta’s Llama 2. Then it launched Sarvam-2B in mid-2024 as a lightweight model trained from scratch for efficient Indic processing. By October 2024 it had shipped Sarvam-1, a 2 Bn parameter model spanning 10 major Indic languages and English. In 2026 it moved upmarket with Sarvam-30B and Sarvam-105B, open-weight models designed for tougher enterprise reasoning and longer context across more than 22 local languages.
The business is already live. Sarvam’s systems are deployed across banking, insurance, government, and defence. Tata Capital is one of its named customers. Conversational agents make up nearly 80% of ARR, which has been estimated at about $12 Mn. The startup’s conversational platform now handles more than 2 Mn interactions a day, while its inference platform processes over 10 Mn API calls daily. In a June 16, 2026 interview, Vivek Raghavan said voice AI alone had crossed 2 Mn calls a day and API usage had climbed to more than 300 Mn calls in the last month. FY26 revenue stood at ₹45.1 Cr, up sharply from ₹1.5 Cr in FY25.
Sarvam has also been willing to chase unusual infrastructure bets. Earlier in 2026, it partnered with spacetech startup Pixxel to build Pathfinder, an orbital data centre satellite targeted for launch in late 2026. It sounds a little wild. But it fits the company’s broader view that AI advantage won’t come from apps alone.
Fundraising details
This round made Sarvam India’s 130th unicorn. It raised $234 Mn as part of a $300 Mn Series B at a post-money valuation of $1.5 Bn. HCLTech led the round, with Bessemer Venture Partners, Khosla Ventures, and Peak XV Partners also participating. HCLTech’s exchange filing said it acquired 41,421 equity shares, equal to a 10.46% stake, for ₹1,427.3 Cr.
The money is earmarked for research and compute. It’s also meant for rollout. Sarvam plans to build next-generation models centered on agentic AI, coding, and cybersecurity, expand compute infrastructure, and scale deployments across sectors. The founders have also said the round should help with global hiring and productisation, including plans connected to a San Francisco office.
How does Sarvam compare with rivals?
Sarvam’s competition comes from a few different camps. Krutrim is pushing open models and giant compute ambitions, including plans around a large supercomputer. Gnani is strong in enterprise voice AI and multilingual speech. Another IndiaAI Mission contender, Soket, is building a 120 Bn parameter open-source foundation model aimed at areas like defence, healthcare, and education.
But the older alternative usually isn’t another sovereign-model startup. It’s the patchwork enterprise stack: a global LLM on top, separate speech and translation vendors underneath, then a services integrator trying to stitch it all together. Sarvam’s edge is that it can sell the whole bundle — models, Indian-language speech, document intelligence, APIs, and deployment. It’s also arguing that more of the compute and value chain should stay in India. That’s the strategic bet HCLTech is backing.
Why does the Sarvam AI funding round matter?
This isn’t just a vanity valuation story. HCLTech gives Sarvam something most frontier-model startups don’t get early: serious enterprise distribution and instant credibility with cautious buyers. It also gives Sarvam a partner that already sits inside the banking, insurance, and public-sector accounts it wants. Because HCLTech is taking a 10.46% stake, this looks like a strategic alliance, not a ceremonial check.
The round also marks a shift in what Sarvam is trying to become. It’s no longer enough to be the promising Indian-language model company. Sarvam now wants to ship larger models and build enterprise-grade agentic systems. It also wants to widen Samvaad access through self-serve onboarding, free usage tiers, and usage-based pricing for startups, developers, and SMBs. That’s a move from bespoke enterprise selling toward something more platform-like.
There’s risk baked into that ambition. Frontier-model R&D burns cash fast, and Sarvam is stretching into coding and cybersecurity while still scaling voice, API, and document products. If it can turn HCLTech’s channel into repeat deployments, this round could matter a lot more than the unicorn label itself.
How big is the market Sarvam AI is chasing?
The near-term market here isn’t “all AI.” It’s enterprise generative AI in India — especially jobs where voice, documents, multilingual support, and compliance all collide. One forecast puts India’s enterprise generative AI market at $183.4 Mn in 2024 and $1.225 Bn by 2030, which works out to a 38.3% CAGR from 2025 to 2030.
And the mix of that market matters. Software was the biggest revenue segment in 2024, but services are expected to grow faster. That lines up with what Indian buyers often need in practice: not just a model endpoint, but deployment help, workflow integration, and ongoing tuning before AI is useful inside a bank, insurer, or government department.
The macro tailwind is clear. Indian enterprises want lower-cost AI and better support for local languages and scripts. They also want more control over where data and compute live. Policymakers want sovereign model capacity for roughly the same reasons. Sarvam didn’t invent that demand — but it showed up at the right time to package it into a product and an investment story.
Final take on Sarvam AI funding
Sarvam has gone from an Indic AI research bet to a $1.5 Bn company in under 3 years, which is fast even by 2026 AI standards. But the hard part starts now. The Sarvam AI funding story gets more interesting from here: whether wider Samvaad access, bigger compute, and HCLTech’s enterprise reach can turn strong usage into durable revenue instead of just bigger expectations.
Read how Foodstories raised ₹50 crore from Nikhil Kamath and Narotam Sekhsaria Family Office to expand its omnichannel gourmet food platform that blends premium grocery retail, food discovery, and fast delivery across India's major cities.
FAQ
- What is the latest Sarvam AI funding round? As of June 16, 2026, Sarvam had raised $234 Mn in the first close of a $300 Mn Series B at a $1.5 Bn post-money valuation. HCLTech led the round with a $150 Mn commitment, and Bessemer Venture Partners, Khosla Ventures, and Peak XV Partners also joined.
- How does Sarvam AI work? Sarvam works as a full-stack enterprise AI platform for Indian languages. Companies can use its APIs for speech, translation, chat, and document digitisation. They can also deploy Samvaad agents that connect to internal systems, act on customer requests, and run across voice, WhatsApp, and web from one setup.
- Who are the founders of Sarvam AI? Sarvam was founded by Vivek Raghavan and Pratyush Kumar in 2023. Raghavan is known for work tied to India’s digital public infrastructure, while Kumar built deep research credentials through IIT Bombay, ETH Zurich, IBM, Microsoft Research, and Indian-language AI work before starting Sarvam.
- Is Sarvam AI a chatbot company or a foundation-model startup? It’s closer to a full-stack enterprise AI company than either label alone. Sarvam builds foundation models such as Sarvam-30B and Sarvam-105B. But it also sells production products like Samvaad and APIs for speech, translation, and document intelligence — which is why its revenue is increasingly tied to enterprise deployment, not just model access.
