Daybreak - How India became the world's biggest AI lab, and not an architect

Episode Date: December 16, 2025

India has the engineers, the users, and the ambition to be an AI superpower. But as OpenAI floods the market at ₹399/month, Google invests $15 billion, and global giants harvest Indian data, a critical question emerges: Will India settle for being the world's largest AI user, or can it become a builder that matters?From DeepSeek's $6M shock to the race for AI sovereignty, we connect the dots on India's AI moment—and what could be next.Tune in. Episodes mentioned: Deepseek: Spotify | Apple | Youtube ChatGPT 399 Plan: Spotify | Apple | YoutubeIndia's Sovereign AI: Spotify | Apple | YoutubeDeloitte's AI blunder: Spotify | Apple | YoutubeAI Browsers: Spotify | Apple | YoutubeWhy AI minds are refusing big bucks: Spotify | Apple | YoutubeCall Centres are being rewritten by AI: Spotify | Apple | YoutubeWrite to us with your thoughts at [e-mail]! 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 Ganesh, 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 same. 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 managed 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.

to 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 an alert, as soon as we release our first 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.

Just last week, Time magazine's cover image for Person of the Year leaked. Before the leak, there was, of course, a lot of speculation about who it could be. Zoran Mamdani, Pope Leo, Trump, Taylor Swift, all the notable headline makers of the year were named. But when the cover dropped, it wasn't a single person. It was a group of people all seated in a single line. A callback to the iconic lunch atop a skyscraper picture from the 1930s. If you haven't seen it, it's a group of construction workers seated on a steel beam during the lunch hour as New York City sprawls into the distance behind them.

This time though, Times cover didn't feature construction workers. It was tech leaders instead. Or, as Time called them, the architects of AI. It features several recognizable figures we all know. Mark Zuckerberg, Elon Musk, Jensen Huang, Sam Altman, Demis Hasabas, who's the CEO of Google's Deep Mind Division, Dario Amode, the Anthropic CEO, along with AMD's Lisa Su and Stanford Professor Fei-Fei Lee,

who's also called the godmother of AI. As expected, Times shorted. has sparked some heated discourse online. Some people thought it was in bad taste to imitate a picture of working-class labourers who risk their lives for daily wages with ridiculously wealthy CEOs. Jimmy Kimmel even called them the eight dogs of the apocalypse. Others took issue with whom the cover seemed to celebrate as the so-called architects. Should it be the CEOs of hyperscalers?

Or the researchers and engineers who actually built the technology? But no matter what side of the debate you fall on, or if you have an entirely different point of view of your own, Times' decision to feature these leaders is a clear indication of one thing. 2025 marked the year AI broke out of the realm of hypothetical, lofty ideals, and entered real-world implementation and large-scale adoption. And daybreak has reported extensively on several of the significant moves AI has seen, especially in India. So, as most of you know, we are a daily news podcast. Right now, you're listening to the 643rd episode.

That's a lot of episodes. Out of those, around 250 of them have been in this year alone. And again, out of those 250 episodes, 15 of them have been on AI. Interestingly, 11 of these 15 were actually just over the last six months or so, which goes to show how much AI development had picked up pace in that time frame. All of this to say, we at Daybreak have been on top of all these stories, and not just a groundbreaking ones, but also events that didn't get as much fanfare.

More importantly, we've also been quite dedicated to exploring how global developments in AI affect us here in India. And in this episode, we'll connect the dots across the biggest AI stories of the year to help you understand what actually changed and why it matters going forward. Welcome to Daybreak. a business podcast from the Ken.

I'm your host, Rachel Bergeese, and every day of the week, my co-host, Nickas Sharma and I, will bring you one new story that is worth understanding and worth your time. Today is Wednesday, the 17th of December.

Now, one of the first big industry-shaking events we had in this year was the launch of China's deep seed. Up until then, the larger narrative around AI had been clear. While the use of AI was for everyone, the engineering of this technology was for a select few. because the three pillars of AI models and training were all expensive resources,

compute, data and the capital necessary to fund the previous two. And the more you have of all three, the faster your AI models can develop. But China's deep seek proved that this narrative was not a universal truth. This smaller startup with about 140 engineers, mostly recent university graduates, had made a model called Deepseek R1 that was comparable in capability to open. open AI's O1 model. And they had made it at a fraction of the cost. On top of having great capability for the price it was built with,

it was also open source, which means basically anyone could download it and build on top of it. DeepSeek basically became a ray of hope for Indian AI startups and the larger state-run mission. You see, all the building blocks of AI are things India was low on, compute capacity, data, capital, but deepseek was proof that cheaper yet clever models were possible and even a reality. It opened up all sorts of new possibilities and Indian AI companies jumped on it.

But before they could begin to catch up, Open AI had made a big move into India. More on this in the next segment. In August this year, OpenEI decided to sell Chad-GTP-Go in India for $399. Then in November, they decided to sell. to offer it to users for free for a whole year. This was the cheapest plan in the world. In an episode covered by my co-host Snicktha, she spoke about how this move had less to do with finding

any sort of paying consumer base in India. Because we're kind of notorious for being cheap, not really a crowd that's going to pay big bucks for a program that also has free variants. So one reason is obviously to build habits now to win wallets later. The second is the sheer size of India's consumers. market and the wealth of data that comes with it. But if India's smart, tech savvy, English-speaking consumers are feeding these international

companies with data, what happens to our local players? As we discussed earlier, data is everything when it comes to LLMs. Every prompt, every rejection of a prompt, even every, that works, thank you, becomes more data for the LLM to feed and grow on. And of course, these companies are not sharing their. India made datasets with any other players. Now, where does that leave local players? So, players like Servam AI and Bhavisha Garwal's crew trim

are building multilingual Indic models that will have very specific use cases in India and for India-specific problems, especially in sectors like agriculture, local compliance and Kirana logistics. The India AI mission is also set to launch its sovereign LLM called Bharat Jens in February 26. That's just a couple months away.

It claims to be ethical and inclusive with more than 20 languages and rooted in Indian cultural context. The mission as a whole has also promised 10,000 crore rupees to build out AI infrastructure in India. So, local players are definitely trying to catch up and build models that stand on their own. And as long as they cater to Indians with Indian problems, they'll still have a chance to make themselves indispensable, but only low. Now, of course, everyone knows that India is highly unlikely to make the next open AI happen. So why invest so heavily in local infrastructure? Stay tuned. Well, turns out, sovereign AI is more of a necessity than a choice.

It's the complex answer to a simple question, which is, who should decide what the AI in your country knows, how it behaves, and whose interest it serves. Now, as more and more business and tech infrastructure starts to rely on AI, the more dangerous it becomes to let foreign bodies control it. Take China, for instance. Even though it maintains a majorly protectionist tech environment, it still relies heavily on U.S. foundation models. Chinese researchers refer to it as a strategic vulnerability.

Their biggest concern? If the U.S. decides to tighten export controls, cap API access, or restrict foundation model licensing, Chinese industries from finance to health to manufacturing could be left scrambling. But even with these obstacles, countries like China and South Korea still show how AI sovereignty works. South Korea's Hyper Clova X model dominates Korean benchmarks, backed by 300,000 GPUs deployed in two years. China, despite its scale, does face a little bit of a hard reality. In public, it phrases its sovereign models. But in secret, they rely on US models because domestic alternatives lag years behind due to chip restrictions and training costs.

On the other hand, while India has the ambition, the talent and the market, it lacks infrastructure. It's a massive AI consumer without the GPU capacity to train competitive models. And its sovereign model target arrives 18 months behind Korea. There's also another reality that complicates this. See, one of India's largest exports is cheap labor and services. AI fundamentally weakens that advantage. So, as we monetize our call center and banking services data and train AI to take over these jobs, we are losing that advantage as well.

Plus, we're constantly selling our resources to foreign players. Google is investing billions in an AI hub in the country. Open AI is setting up in KUA, its first region-specific benchmark designed to test the performance of AI models in understanding Indian cultural nuances and languages. Anthropic is all but ready to open an office in Bangalore. All these companies have identified India as the second largest market for AI in the world, and they have no lack of capital to take whatever resources they need. Short of building a firewall for foreign tech, there's really not much India can do to protect the interest. of local players. So while Open AI, Google and Anthropic are rapidly filling the gap, India risks becoming a leading AI market without being a builder. And what we are left with is a

future that's built for India but not by India, where foreign models become too entrenched for domestic alternatives to catch up. That's it for today. Sure, it's a bit of a dismal outlook as far as Indian AI is concerned, but like I mentioned earlier, there's still hope for them to build use cases locally. And now, quickly before we end, here are some other highlight episodes from the year. First, in an episode on Deloids Gen AI Blunder in an Australian government report, we discussed how consultancies are rewriting the importance of human certainty through AI. Secondly, in another episode on AI browsers, we explore what and who AI browsers. and extensions are really far.

Third, despite some real big check offers from AI giants, the tech industry's brightest minds are turning them down for something better. What could that be? Fourth, as I mentioned the monetizing of call center data earlier, we cover that on daybreak as well in an episode on how call center employees are essentially training AI to take over their jobs eventually. And that's a wrap. We'd love to hear from you what you think

were the biggest events in AI this year. And even if you think there are any major fails. At Daybreak, we're excited to cover both sides of the coin in 26. And we hope that you'll be here to listen to us. 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 offers daily long-form feature stories, newsletters, and a whole bunch of

premium podcasts. To subscribe, head to the ken.com and click on the red subscribe button on the top of the Ken website. Today's episode was hosted and produced by my colleague Rachel Vargis and edited by Rajiv Sien.