Daybreak - If AI is expensive everywhere then why is ChatGPT cheapest in India?

Episode Date: August 20, 2025

OpenAI has just launched a special, India-only plan for ChatGPT. It makes access cheaper here than almost anywhere else in the world. At first, it looks like a win for millions of Indians who’ll get to try cutting-edge AI at a fraction of the cost.But for India’s AI startups, it may be a different story. Competing with global giants that have billions in capital, access to compute, and a head start on scale is already tough. Add aggressive pricing, and the playing field gets even steeper. After all, if Indians aren’t paying with money, they’re likely paying with something else: data and usage.So where does that leave Indian AI? Can startups like Sarvam, Krutrim, and others still carve out a niche through language, verticals, or local trust or will they be reduced to distributors for the biggies?Find our last episode on Sarvam here. Want to join The Ken's team? Fill this form.Daybreak is produced from the newsroom of The Ken, India’s first subscriber-only business news platform. Subscribe for more exclusive, deeply-reported, and analytical business stories.

Hi, this is Rohan Dharma Kumar. If you've heard any of the Ken's podcasts, you've probably heard me. My interruptions, my analogies and my contrarian takes on most topics. And you might rightly be wondering why am I interrupting this episode too. It's for a special announcement. For the last few months, I and Sita Raman Ganeshan, my colleague and the Ken's deputy editor, have been working on an ambitious new podcast. It's called Intermission.

We want to tell the secret sauce stories of India's greatest companies. Stories of how they were born, how they fought to survive, how they build their organizations and culture, how they manage to innovate and thrive over decades, and most importantly, how they're poised today. To do that, Sita and I have been reading books, poring over reports, going through financial statements, digging up archives, and talking to dozens of people. And if that wasn't enough, we also decided to throw in video into the mix. Yes, you heard that right. Intermission has also had to find its footing in the world of multi-camera shoots in professional studios, laborious editing, and extensive post-production. Sita and I are still reeling from the intensity of our first studio recording.

Intermission launches on March 23rd. To get alert, as soon as we release our first video. episode, please follow intermission on Spotify and Apple Podcast or subscribe to the Ken's YouTube channel. You can find all of the links at the ken.com slash I am. With that, back to your episode. For decades after its independence, India told itself a story about technology. That big technology was not for us. Instead, the country needed appropriate technology, small, local, targeted. Machines designed not to transform society, but to solve narrow immediate problems. The Indian state encouraged specific solutions over sweeping revolutions.

Now, fast forward to today, the world is in the middle of an AI race. And Open AI, the most influential player of them all, has just rolled out its cheapest subscription anywhere in the world, not in the US, not in Europe, but in India. 399 rupees a month. That is Chad GPD Go that was launched just this Tuesday. It is a price that makes no sense if you really think about the costs. Running these AI models burns the same GPUs, the same electricity, whether it's in Bangalore or in the US.

Computation does not get cheaper in India. So why should Indians be paying less? The answer lies in strategy. India is once again the testing ground, the laboratory, the place where millions of students, freelancers, small businesses, can stress-test these tools, feed them data, and more importantly, form daily habits around them. And that raises a much bigger question.

Because if India becomes once again the market that supplies global tech giants with users and usage data, then what happens to Indian startups? Can companies like Zavam AI, Crotrim and Kisaniai build anything meaningful when open AI and perplexity are already flooding our market with cheap or free tools? Welcome to Daybreak, a business podcast from the Ken. I'm your host, Nidda Sharma, and I don't chase the news cycle. Instead, every day of the week, my colleague Rahil Filippos and I will come to you with one business story that is worth understanding and worth your time. Today is Thursday, the 21st of August. When the internet and social media were exploding globally, India became the biggest growth market for them.

Meta, Google, YouTube, all of them built enormous user bases here. But India did not build its own equivalence. No Indian Facebook, no Indian Google, because the economics of scale, data and capital all tilted the game in favour of these global giants. Now, the same dynamics are showing up in the AI race. Take Open AIS 399 rupees plan or perplexity which is literally free if you are an Airtil subscriber in India. Both are designed to maximize usage, not revenue for these companies, because the strategy is clear. Win habits now and monetize later. You see, for global AI companies, India is a gold mine.

nearly a billion internet users, a young English-speaking base, millions of students, professionals, freelancers, all hungry for productivity tools. In other words, it is the perfect environment to harvest usage data at scale. And usage is everything in AI. Every prompt, every correction, every follow-up query becomes training data. And that data feeds back into the model and makes it better and widens the gap between the incumbents and everybody else. And that is why this wave of AI adoption is dangerous for Indian startups.

They're not just competing on features. They're competing against OpenAI's feedback loop of global scale, better models and unlimited capital. In fact, OpenAI CEO Sam Altman not long ago put it quite bluntly. It was, he said, totally hopeless to compete with them on foundation models. hopeless, and the economics back him up. Open AI has raised close to $18 billion to date. Meta's training costs vary widely. Training a model like Lama 270B costs a few million dollars,

while its largest Lama 3.1 model reportedly cost around $170 million. Now, compare that with India. Crotrim, OLA founder Bhawe Shagarwal's AI venture, raised $50 million in its debut round to become a unicorn. Most Indian startups like Kisaniai build on top of open source models such as Metas Lama or Mistral rather than training full foundation models from scratch. And then there are three more brutal realities. First, no firewall. For example, in China, companies like Baidu and I Fly Tech thrived because U.S. platforms were blocked.

India offers no such protection. Chad GPT, perplexity, Gemini all operate freely over here. The second is compute costs. Running and training AI models costs the same everywhere in the world. There is no India discount on Nvidia GPUs. And third is the capital gap. Building big AI takes billions in compute and research. Indian VCs simply do not write checks of that size.

Without sustained state or corporate banking, scaling is nearly impossible. And the result is India, once again, risks becoming the global lab. the place where usage is abundant, but ownership of the technology is actually somewhere else. And yet, the story is not over. Because when Open AI and perplexity tighten their grip on the user side, India has its own counterweights on the supply side. Stay tuned. Let's take Sarvam AI, for example.

It is the most ambitious attempt yet to build an Indian Foundation model company, founded by ex-Microsoft Research Scientist, supported by N. Nandan Nelikani and backed by the top VCs of the country, Sarvam pitches itself as the multilingual Indic first alternative. Its models were built in record time, less than a year. Sarvam also has heavyweight partners like Microsoft, Meta and Nvidia. But even with all of this support, products have been slow to launch. Capital is finite and in an industry where OpenAI is burning billions, Sarvam is still a baby. Its models are trained not just on English internet data, but on synthetic data sets for coding, reasoning and Indian languages. Because real-world

Indic data is scarce and messy and Sarvam generates synthetic data to fill these gaps. Also, Sarvam has released a whole host of multilingual products from AI agents, voice and text models to a workbench aimed at legal professionals even. Enterprise customers who use Sarvam's services are satisfied with the performance of its products. But developers have flagged issues with its voice-based models. And even its text model is primarily trained on synthetic data, which could lead to nonsensical answers if left untested. We actually covered this in detail in another episode of Daybreak,

and I will link it to the show notes of this one. Now, let's move to Crotrim. Bhawe Shagarwal's new bet, which is after Ola. It brands itself as India's first multilingual LLM. Still in its early days, but with money raised and an ecosystem building, it wants to stand alongside Sarva. And then there are some other smaller players, like Kisaniai, which is building language models for farmers,

or Korova, which launched Bharat GPD to capture Indian linguistic nuances. Each of them are tackling niche Indian problems, and all of them are running into the same choke points, limited capital, scarce data and the high cost of GPUs, which brings us to the government. Earlier this year, IT Minister Ashwini Weishnav announced the India AI mission, a more than 10,000 crore-rupy program and the goal is to build AI domestic capacity. That means foundational models, public data sets, and most critical computing power. Ten companies, from Yota to Geo to Tata Communication and some others,

will together supply nearly 19,000 GPUs. Almost half of them are from Yota alone. Now, these GPUs will be pooled in as public infrastructure, so startups don't have to buy them outright. Now, this is pretty significant because compute, like I told you earlier, is the oxygen of AI. Without GPUs, you cannot. train or scale. The government also promises to support startups directly with grants, procurement,

and partnerships. The pitch is that AI models aren't just commercial assets but strategic resources as important to sovereignty as telecom or energy. So now the question is, is this enough to tilt the playing field? The answer depends on whether startups can carve out moats in three specific areas. First, localization. India's messy diversity, language, dialects, regulations is hard for global players to prioritize. Startups that can tune AI to these realities will stand out. Second is vertical AI. Specialized tools for industries like Kirana logistics, agri-supply chains, or compliance even. Open AI will not optimize for the Indian farmer or for small-town accountants, but a local startup can.

And the third is integration. Embedding AI into the workflows of SMEs or small or medium enterprises, schools or government services. In India, trust and relationships often matter more than raw model quality. Startups that can build on trust can carve durable niches. But the big risk remains. If Indian startups fail to build these modes, they will end up as resellers or distribution partners for global giants.

This is similar to what happened in the internet era, when most Indian startups built on top of Googles and Meta's rails, but never really challenged them. So, the final takeaway is this. India's AI startups do have a chance, but not at becoming the open AI of India. That playbook is closed. Their real shot actually lies in being indispensable locally

while the giants fight globally. Otherwise, history is going to repeat itself. India will become the world's favorite user base for AI. Millions of prompts, billions of interactions, but the real value capture, which is the capital, the intellectual property, the influence will flow somewhere else. And that is the question that is hanging over India's AI moment. Will we settle for being the market or will we build?

Even if we build different. Daybreak is produced from the newsroom of the Ken, India's first subscriber-focused business news platform. What you're listening to is just a small sample of our subscriber-only offerings. A full subscription unlocks daily long-form feature stories, newsletters and podcast extras. To subscribe, head to the ken.com and click on the red subscribe button on top of the Ken website. Today's episode was hosted by Snigda Sharma and edited by Rajiv CN.