AI Startups in India: Who Has Shipped, and Who Hasn't

Most lists of AI startups in India are lists of funding rounds. That ranks the wrong thing. India's first AI unicorn has spent 2026 retreating from the models it was funded to build, and the largest "round" of the year is half debt that had not closed when announced. So this list is ordered by what each company has actually put in users' hands, with the money as context — and, for the three we have tested ourselves, with prices in rupees.

AI startups in India by funding, for reference

Inc42's Datalabs tracks over 1,500 AI startups and ranks the most funded "strictly by total funding raised to date". Its top 20 for 2026, H1 figures, opens:

Rank	Startup	Total funding (Inc42)	Founded
1	Ola Krutrim	$524M	2023
2	Wispr Flow	$361.6M	2021
3	Sarvam AI	$275M	2023
4	Emergent	$230M	2025
5	Kore.ai	$220M	2013
6	Brahma AI	$150M	2023
7	Ema	$138M	2023
8	Freehand	$100M	2024
13	Neysa	$49.9M	2023

Two cautions before reading it. Inc42 counts Indian-founded companies, and several of these are headquartered abroad. And the Neysa figure predates a Blackstone-led financing of up to $1.2 billion announced in February — which, as explained below, is not $1.2 billion of equity.

1. Sarvam AI — the one that shipped a frontier-class open model

Bengaluru, founded August 2023 by Vivek Raghavan and Pratyush Kumar, both formerly of AI4Bharat at IIT Madras.

What it shipped: Sarvam-30B and Sarvam-105B, announced at the India AI Impact Summit on 18 February 2026 and open-sourced under Apache 2.0 in early March. The 105B model card describes a mixture-of-experts model with 10.3 billion active parameters, a 128K context window, and "state-of-the-art performance across 22 Indian languages" for its size. A consumer assistant, Indus, launched in beta on the larger model. MeitY selected Sarvam for the IndiaAI Mission in April 2025, with ₹246.72 crore in compute and financial support per PIB.

The money: on 15 June 2026 Sarvam announced a $234 million first close of a $300 million Series B at a $1.5 billion post-money valuation, led by HCLTech with $150 million, with Bessemer joining Khosla Ventures and Peak XV. TechCrunch called it "India's newest AI unicorn". Inc42 later reported a ~$74 million extension with NVIDIA participating, described as approved by the board rather than closed.

Usage figures are company-reported: 2 million conversational interactions a day, 10 million API calls, 17 million farmers reached through voice agents for the agriculture ministry. Revenue was not in the announcement; Business Standard, citing unaudited figures HCLTech shared, put FY26 revenue at ₹45.1 crore against ₹1.5 crore the year before.

2. Neysa — the GPU landlord

Mumbai, founded 2023, co-founder and CEO Sharad Sanghi, about 110 employees.

What it shipped: compute. Roughly 1,200 GPUs live, with a target of "more than 20,000 GPUs over time", per TechCrunch's report of 15 February 2026. Neysa joined the IndiaAI Mission's compute pool in 2025.

The money, read carefully: Neysa's own press release describes "equity capital of up to $600 million" from Blackstone and co-investors — Teachers' Venture Growth, TVS Capital, 360 ONE Asset, Nexus — "on the basis of which Neysa intends to secure an additional $600 million of debt financing, subject to documentation." Blackstone takes a majority stake; India's competition regulator cleared it in May. So the headline $1.2 billion is half equity and half debt that was not confirmed closed. Before this, Neysa had raised about $50 million, which is the figure in Inc42's table.

3. Gnani.ai — voice, in eleven languages

Bengaluru, co-founder Ganesh Gopalan. One of the four startups picked in the IndiaAI Mission's first phase, with ₹177.27 crore in support per PIB.

What it shipped: Vachana STT, a speech-to-text model trained on more than a million hours of voice data, covering Hindi, Bengali, Tamil, Telugu, Kannada, Malayalam, Marathi, Gujarati, Punjabi, Odia and Assamese, and Vachana TTS with zero-shot voice cloning across 12 Indic languages. The voice-to-voice model is 5 billion parameters, with a 14-billion flagship planned. Accuracy claims are the company's.

The money: a $10 million Series B in March 2026 led by Aavishkaar Capital with Info Edge Ventures, after a $4 million Series A in 2024. Small next to the names above, which is the point: a shipped model on a tenth of the capital.

4. Krutrim — the cautionary tale

Bengaluru, Bhavish Aggarwal's AI company, India's first AI unicorn in January 2024.

What happened: on 5 May 2026 TechCrunch reported that Krutrim was "shifting from AI model development to cloud services". A late-2025 overhaul paused chip design, more than 200 roles were cut according to local reports, and the Kruti assistant app was pulled from app stores in April. The company reported FY2026 revenue of about ₹3 billion, a first annual net profit and 25-plus enterprise customers — and did not answer TechCrunch's questions on its revenue mix, after earlier reports put about 90% of FY2025 revenue from group companies. Analyst Sanchit Vir Gogia's line: "the standard of proof must rise with the claim."

The money: Inc42 ranks it first at $524 million. In February 2025 Aggarwal committed ₹2,000 crore and pledged ₹10,000 crore by 2026; Krutrim denied an April 2025 report that it was seeking outside capital. Whether the pledge was met is not reported.

It remains the most-funded AI startup in India and, by what it has shipped to users this year, the least.

5. BharatGen and the rest of the IndiaAI Mission twelve

The IndiaAI Mission selected four startups in its first phase — Sarvam, Soket AI, Gnani and Gan AI — and twelve in total, including BharatGen, an IIT Bombay-led consortium, plus Avataar AI, Zenteiq, GenLoop, Intellihealth, Shodh AI, Fractal Analytics and Tech Mahindra's Maker's Lab, per PIB. The mission's compute pool stands at over 38,000 GPUs at ₹65 an hour.

BharatGen's Param2, a 17-billion-parameter model supporting 22 Indian languages, launched at the February summit with the largest allocation, ₹1,058.52 crore. Soket's planned 120-billion-parameter open model and Gan AI's 70-billion text-to-speech model were announced as plans; as of 11 October 2026 we found no launch for either.

6–8. The legal-AI startups we have actually tested

This is where this list differs from the others. We reviewed seven AI tools for Indian legal work, with published rupee pricing, for our legal drafting guide. Three stood out:

CaseMine — Gauge Data Solutions. Our highest-rated Indian legal tool at 4.3: research across the Supreme Court, all 25 High Courts and major tribunals, and the one that flags overruled judgments. ₹1,999 a month, with free plans for students, prosecutors and judges.

Niyam.ai — 4.2. Inline citations to the primary source on every answer and a good-law check on every cited case. ₹1,500 a month, rupee-only billing.

Prism AI — from Indian Kanoon, built on its 30-million-judgment corpus. 4.1, with a permanent ₹0 plan and the cheapest paid entry at ₹5,000 a year plus GST.

None of these appears on funding lists, and all three have more paying-user-facing product than several companies that do. The wider field is on our AI tools for legal work page.

9. Miko — consumer hardware from Mumbai

Emotix, founded by three IIT Bombay alumni, makes the Miko companion robots for children aged five and up, sold in the US at $299 for the Miko 3 and $199 for the Mini, with Indian marketplace prices to verify at the time you buy. It is the rare Indian AI startup whose product is a physical thing in homes; we covered it in AI for kids.

How to read the numbers

Four habits, learned writing this list of AI startups in India:

1. Equity and debt are not the same headline. Neysa's $1.2 billion is $600 million of each, and the debt was "subject to documentation".
2. "Set to raise" is not "raised". Sarvam's extension was a board resolution when reported.
3. Company revenue is unaudited until it isn't. Krutrim's profit and Sarvam's ₹45.1 crore both came without audited accounts.
4. Funding trackers count Indian-founded, not India-based. Check the headquarters before calling a company an Indian startup.

FAQ

Which is the most funded AI startup in India?

Ola Krutrim, at $524 million by Inc42's count, followed by Wispr Flow at $361.6 million and Sarvam AI at $275 million — a figure that predates Sarvam's Series B extension. Neysa's Blackstone-led financing of up to $1.2 billion is half debt and is not reflected in those totals.

Is Sarvam AI a unicorn?

Yes. On 15 June 2026 it announced a $234 million first close of its Series B at a $1.5 billion post-money valuation, led by HCLTech, per its own announcement and TechCrunch.

Which startups are building India's sovereign AI models?

Under the IndiaAI Mission, the first four were Sarvam AI, Soket AI, Gnani and Gan AI, expanded to twelve including BharatGen, Avataar AI, Zenteiq, GenLoop, Intellihealth, Shodh AI, Fractal Analytics and Tech Mahindra's Maker's Lab. Sarvam, BharatGen and Gnani launched models at the February 2026 summit.

What is happening with Krutrim?

In May 2026 it shifted from building AI models to selling cloud services, after pausing chip design and cutting more than 200 roles, per TechCrunch. It reports a first annual profit but has not disclosed how much revenue comes from Ola group companies.

Which Indian AI startups have products I can use today?

Sarvam's Indus assistant and its open models on Hugging Face; Gnani's Vachana speech models; CaseMine, Niyam.ai and Prism AI for legal research, each with published rupee pricing; and Miko's robots. Soket and Gan AI had announced plans but no launch we could find.