Open the box of a new smartphone in India, and you are greeted by a familiar landscape. The icons for Google, WhatsApp, and Instagram are laid out like a digital welcome mat. The setup process prompts you to choose a language, a nod to the country’s diversity. But for decades, the deepest, most intelligent functions of the device, especially its voice assistants and AI, have spoken a global language first and an Indian one as a hesitant, often clumsy, second. The core intelligence has always felt imported.
That might be starting to change. In a move that has flown somewhat under the radar amidst the noise of big funding rounds and IPO filings, Finnish phone maker HMD has quietly placed a significant bet on a homegrown Indian AI. Its newest device, the Vibe 2 5G, comes with something different pre-installed: the Indus chatbot, built by the Indian AI company Sarvam. This is not just another app bloating the home screen. It represents a potential turning point, a test case for whether a truly Indic language model can offer a compelling enough reason for a user to choose one piece of hardware over another.
For years, we have talked about building for Bharat. We have seen it in fintech with UPI, in ecommerce with vernacular interfaces, and in content with regional OTT platforms. Now, the frontier is artificial intelligence. The question is no longer if AI will reach the masses, but how it will speak to them. With this partnership, HMD and Sarvam are suggesting the answer is simple: it will speak their language.
A Symbiotic Bet on Differentiation
In the brutal, wafer-thin margin world of the Indian smartphone market, differentiation is everything. For HMD, a company carrying the legacy of Nokia and trying to carve out a space against Chinese behemoths and Korean giants, competing on specs alone is a losing battle. This is where the partnership with Sarvam becomes a clever strategic play.
Instead of just another Android skin or a slightly better camera sensor, HMD is offering a unique software experience rooted in Indian identity. The Indus app is more than a feature; it is a statement. It tells the potential buyer in Lucknow, Coimbatore, or Bhubaneswar that this device was built with their conversational reality in mind.
For Sarvam AI, the challenge is the inverse of HMD’s. They have the deep tech, but they lack distribution. Building a foundational AI model is an immense undertaking, but getting it into the hands of millions of users is arguably a greater challenge. The Google Play Store is a graveyard of brilliant apps that never found an audience. A pre-installation deal with an Original Equipment Manufacturer (OEM) like HMD is the holy grail for a consumer-facing AI startup. It bypasses the brutal user acquisition (CAC) game and puts the product directly into the user’s journey from the moment they unbox the phone.
This is a classic ecosystem partnership where each party provides what the other desperately needs. HMD gets a unique selling proposition, and Sarvam gets a massive, zero-cost distribution channel and an invaluable real-world laboratory to test and refine its models.
The Technology of Talking Indian
So, what makes the Indus chatbot different? The answer lies in the data and the design philosophy. Sarvam’s model is not a Western model retrofitted with Indian language packs. It was built in India, for India.
The core of the technology is a 105-billion-parameter model. In the world of Large Language Models (LLMs), the parameter count is a rough measure of an AI’s scale and sophistication. But the real magic is not just the size, but the training. The model supports 22 Indic languages, a feat in itself. However, its most critical feature, and the one that demonstrates a deep understanding of the Indian user, is its ability to handle mid-sentence code-switching.
Anyone who has spent five minutes in a conversation in urban or semi-urban India knows that we rarely speak in pure languages. We flow seamlessly between English and Hindi, Tamil and English, or Kannada and English. A sentence might start in one language and end in another. This is the reality that global AI assistants have consistently struggled with. They are built for linguistic purity, a concept that does not exist on the streets of Mumbai or Bangalore.
Indus is designed to understand this linguistic mix. This ability to comprehend ‘Hinglish’ or ‘Tanglish’ without missing a beat is Sarvam’s core bet. It’s a subtle but profound difference that could make interacting with the AI feel natural and intuitive, rather than a stilted translation exercise.
The partnership, first announced during the India AI summit in New Delhi earlier this year, is still in its early days. The Indus app, as it exists today on the HMD Vibe 2 5G, has its limitations. It requires an internet connection to function, putting it out of reach for users in areas with patchy connectivity. Furthermore, it doesn’t yet have deep integration with the device’s operating system, meaning you cannot invoke it with a simple voice command or a hardware shortcut. It exists as a standalone application. But these are early days. The first step is to gauge user appetite. Is a truly local AI assistant a “nice to have,” or is it a killer feature?
The Larger Picture: AI’s Go-To-Market Puzzle
The HMD-Sarvam partnership offers a fascinating glimpse into the future go-to-market (GTM) strategies for India’s burgeoning AI startup ecosystem. While enterprise AI startups can rely on B2B sales teams, consumer AI companies face a much tougher road. This deal could become a blueprint.
Imagine a future where other Indian hardware manufacturers, like Lava, follow suit. Or what if larger players like Xiaomi or Samsung, in their quest for deeper localization, begin partnering with or acquiring Indian AI startups to power their next generation of devices for the Indian market? This single deal could catalyze a whole new category of partnerships.
It also brings up the quiet but intense competition heating up in the background. While Google, Amazon, and Apple have a massive head start with their assistants, their global-first approach leaves gaps that nimble, local players can exploit. Sarvam is not trying to beat Google Assistant at its own game. It is trying to create a new game altogether, one played on the home turf of India’s linguistic diversity.
This trend is not happening in a vacuum. It is part of a broader push, supported by government initiatives like the India AI Mission, to build sovereign AI capabilities. The goal is to ensure that the next wave of technological transformation is not just consumed in India, but also created in India, reflecting the nation’s unique cultural and linguistic contexts. A chatbot on a phone might seem like a small step, but it is a tangible manifestation of this much larger ambition.
The journey ahead for Sarvam and HMD will be telling. They will need to prove that their solution is not just a novelty but a genuinely useful tool that makes life easier for the Indian consumer. The data they gather from these first users will be pure gold, allowing them to iterate on the product, improve its capabilities, and perhaps, eventually, integrate it more deeply into the device. For now, it is a bold experiment. It is a bet that in a country of a billion people and hundreds of languages, the most powerful connection is still the one made when you truly understand what the other person is saying.
