Sarvam Drives 87% of Indic AI’s 2026 Funding in a Market That Is Still Early Stage and Concentrated
India : Tracxn, the global market intelligence platform for private company data, today published its Indic AI: Building AI for Bharat report, a data driven analysis of capital flows into the 34 companies building AI for India’s languages, with data as of 15 October 2026.
The report finds $687M of all-time equity funding across the set, $657M of it since 2021 across 56 rounds. The thread through the data is concentration. A small number of companies and rounds hold most of the capital, and the market is still early, with no late stage round recorded.
A Record Year Built on Two Rounds
Between 2021 and October 2026, companies in the Indic AI ecosystem raised $657M across 56 rounds. 2024 recorded the most rounds of any year at 15, while 2026 recorded the most capital on far fewer rounds. Companies multiplied first, and round sizes grew after. Every round recorded across the period sits at seed or early stage.
Sarvam’s two 2026 rounds were the largest part of the surge, and outside them the set raised $54M across seven rounds. The year also offered a recent illustration of foreign AI dependency, when access to Anthropic’s Fable 5 and Mythos 5 was suspended from June 12 until July 1.
Two Tiers, Two Different Businesses
The report sorts the 34 companies by one test: if the Indic language layer were removed, would the company still have a product to sell? For the 18 companies in the Language is the product tier, such as Sarvam, Gnani, Soket AI and BharatGen, the answer is no. They build speech, translation or language models, or the data to train them.
For the 16 companies in the Language is a differentiator tier, such as Skit.ai, Nuplay AI and GreyLabs, the answer is yes. They sell enterprise products in which Indic language handling is a strength within the product. GreyLabs now serves more than 50 institutions, including RBL Bank, AU Small Finance Bank and IDFC FIRST Bank, one of the clearest signs in the set of paying enterprise demand for Indic voice AI.
Two Unicorns, Two Different Bets
Sarvam and Krutrim were both founded in 2023, have both reached unicorn status and together hold $464M. Sarvam, with $389M raised, is a full stack sovereign generative AI platform and one of the government’s four selected sovereign model builders in the set. Its early language work adapted Meta’s Llama and Mistral’s open models rather than training from scratch.
Krutrim, which has raised $75M, funded its expansion partly through debt collateralised against Bhavish Aggarwal’s own Ola Electric shares. It paused chip design in late 2025, has since shifted from foundation model development toward AI cloud infrastructure and has reported its first profit. It no longer builds Indic AI models and is included in the data for its funding history.
Two Exits, Both Before the Surge
Two companies in the set have exited through acquisition, and both were bought by Reliance group companies in 2019, well before the 2026 funding surge. Reliance Jio acquired about 87% of Haptik, a Mumbai based conversational AI company founded in 2013. The $101M transaction value includes investment for growth and expansion alongside the initial consideration.
Reliance Industrial Investments & Holdings acquired a majority stake in Reverie Language Technologies, a Bengaluru based translation, localisation and speech company founded in 2009, for about $27M. No acquisition has been recorded since, and every round in the period sits at seed or early stage, so the first late stage round is the next marker to watch.
India Stands Alone as the Venture Led Model
Set against Singapore’s state funded SEA-LION programme (S$70M, about $52M), Africa’s LINGUA grants (up to $250K per project across 26 projects) and the Gulf’s state backed HUMAIN in Saudi Arabia and G42 and TII in the UAE, India is the only one of the four built primarily through competing, investor owned companies. Its $687M in equity funding compares with a $77B estimated cost for Saudi Arabia’s HUMAIN project, though the two measure different things.
What happens next turns on two signals. Global model providers are moving into the language gap, with Gemini natively supporting nine Indian languages and OpenAI releasing an Indian language benchmark built with 261 domain experts. And the first late stage round in the set will show whether any company has cleared the threshold from building to scaling.
****
