This isn’t a critique of ElevenLabs or its offerings. In fact, the startup has built one of the most impressive voice-based AI and speech tools available today. Its India chapter is thriving to the point where it is giving Indian AI startups a run for their money.
Even Prime Minister Narendra Modi reposted the AI-dubbed Lex Fridman podcast by ElevenLabs on his X account. And therein lies the critical lesson that many Indian AI startups seem to be overlooking.
Although Sarvam and others also stepped in with dubbing and translation efforts, they were, somewhat surprisingly, not the first choice. This was a crucial opportunity for ElevenLabs to showcase its technology in the Indian market.
Siddharth Srinivasan, who is heading ElevenLabs in India, earlier told AIM that India is ElevenLabs’ biggest market yet, and the company is actively expanding its team in the country to build the future of voice AI, with just 10 people on the team in India right now.
This begs the question: if India wants to make its own AI models and infrastructure with a focus on sovereignty, how is it that a foreign company can still make big bucks from the Indian ecosystem, especially in the same field where India seeks to maintain sovereignty?
That is precisely the question that Sagar Meisheri, lead ML engineer of NLP at Avamo, asked in a LinkedIn post. ElevenLabs’ success in India is not the problem. The real problem lies in this: Indian voices are trained, dubbed, deployed and monetised, yet the money flows back to the US.
And that’s the lesson Indian AI startups have yet to grasp fully.
Why India Still Chooses ElevenLabs?
India is now the number one market for ElevenLabs. The voice AI company reportedly makes around $3 million a month, largely driven by demand from Indian platforms like Pocket FM, NoBroker, Apna, Meesho, and even Star Sports.
ElevenLabs generates revenue through subscriptions for its text-to-speech and voice cloning tools, licensing its default voices, and letting voice actors earn by cloning and selling their voices. For example, Pocket FM has produced over 30,000 hours of AI audiobooks during its experimental phase with ElevenLabs and anticipates tripling its content by the end of this year.
Earlier this month, Meesho also partnered with ElevenLabs to enable real-time multilingual customer support, managing over 60,000 customer calls each day through an AI-powered voice agent.
ElevenLabs states that Hindi is among its top three most used languages, but the company supports 11 Indian languages and plans to expand further through partnerships with the firms mentioned above.
The company recently hosted a hackathon in Bengaluru as part of its global tour, with India recording the highest number of registrations. But why does this happen?
There are startups like smallest.ai, Sarvam AI, Krutrim, Gnani.ai, and a few others that are offering voice models. However, they are far from competing with ElevenLabs when it comes to revenue from India.
Read: Indic AI is Not Inspiring Enough for Indian Developers
But what ElevenLabs does for its developers is exactly what makes it stand out for enterprises. It operates on a SaaS model with two models. In one of them, users can test out the model for free and then scale it up to a subscription model, which costs just $5 a month.
This opens a significant opportunity for the company in the Indian market, as developers can first try it out to gain trust and then figure out if they need it or not. When it comes to others in India, they start with paid APIs or buying the models on the cloud of the provider, making it not so reliant on them.
For example, RevRag, the company focused on building AI agents, still decides to work with open-source and closed-source LLMs from the West. While RevRag is open to deploying homegrown LLMs like Sarvam and Krutrim, Singh notes that latency and production readiness continue to pose challenges.
“We’ve tried Sarvam, but the latency is high. Once it’s sorted, we would love to deploy it because it understands Indian languages better,” he said. As for Krutrim’s Kruti agent, Singh said they haven’t explored it yet as it’s “too early”.
So, What Should be Done?
Indic AI doesn’t have paying customers—at least that is what we think happens. However, if ElevenLabs is making money, it is indeed possible for Indian startups to do the same. But Indian founders are stuck building the next chatbots and AI agents that not many people are using.
Voice AI is everywhere in India, except where it should be: on Indian infrastructure, by the Indian firms.
Read: India has 109 Agentic AI Startups Building in a Vacuum
Meanwhile, the IndiaAI Mission is going strong. The government, under MeitY, has selected startups to focus on building the exact alternatives to what the likes of ElevenLabs, OpenAI, or other Western firms are building. These, notably, are also actively coming to India, and there is no incentive or reason to stop them either.
The IndiaAI initiative has already promised to deploy over 34,000 GPUs across MeitY-backed infrastructure like C-DAC and the IndiaAI Cloud. This would help meet the voice requirements in firms that want to use it after the training is done.
The dream is still unfolding as the firms continue to wait for the GPUs to arrive.
Yet, despite everything, Indian companies still largely place their trust in OpenAI, Google, and Microsoft for their AI capabilities, even if that means spending a little more.
But here’s the disconnect: the taxpayers paid for this infrastructure, Indian researchers trained the voices, and Indian users generate the demand. Yet, the entire value chain is offshored through a US-based API. The effort and the investment happen in India, but the revenue is siphoned abroad.
Meanwhile, in an interview with Entrepreneur India, Srinivasan said the bet is the biggest on the Indian market with partnerships with more AI companies. “We think we are very well placed to be the voice of the Indic internet where content has no barrier and creativity knows no limit,” he told the publication.
A company which is trying this approach is Mukesh Bansal’s Nurix. In a recent video, Bansal revealed a working demo of Nurix’s AI call centre agent in Indic languages. The best part is that he expects the cost of this to soon come down to 40 paise per minute.
But this would also take some time to scale. India must not just be a consumer of AI but also a creator, owner and monetiser of it.
Meanwhile, companies like OpenAI, ElevenLabs, and others from the West will continue to create their strongholds in India—which, to be completely fair, is also not wrong.
[Note: The headline has been amended for clarity.]
