The Rise & Fall of India's AI Dream🔻 | Sarvam AI Case Study

0:00 Economic Times organized an event with open AC was Sam Edman on June 7, 2023 in New Delhi. A famous Indian venture capitalist asked him a question. >> But if you want to build foundational models, how should we think about that? Where is it that a team from India, you know, three super smart engineers with, you know, not 100 million but let's say 10 million could actually build something truly substantial? >> His response, nobody expected. Look, the way this works is we're going to tell you it's totally hopeless to compete with us on training foundation models you shouldn't try and it's your job to like try anyway and I believe both of those things. [laughter] I think it I think it is pretty hopeless but >> an instant outrage forward against Sam Alman in social media but at least so far Sam Alman is right. America has open

0:46 AI, Deep Mind, Anthropic, XAI. China has deepsei, moonshot AI, France has MSL and what about India? If you think we have server, now look at the downloads of the major LLM chart boards in play store. Chart GPD and Gemini literally have 1 billion downloads each. Elon's croc has 100 million downloads. Chinese Deep Seek has 100 million. Anthropic Squad has 50 million downloads. Another Chinese LLM, Kimmy, has 5 million downloads. And finally, our own India's LLM Seron just has around one lakh downloads. So in terms of user adoption, it's a total failure. Now if you look at LLM rankings in Areno.ai, [music] there are seven Chinese models in the top 25 rankings. ZAI, Xiaomi, [music] Moonshotai, Tencent and Deep Sea. There

1:33 are multiple Chinese companies in the race. [music] Whereas India has only one company that is Serom and it is not even in the top 50. So even in the LLM rankings it's a total failure. But the problem is not with failing. We have to encourage failures. Only then we can learn and win. But the problem is with the nationalistic hype [music] and expectations they created. When you call yourself India's sovereign AI, you will be compared with OpenAI, Google, Anthropic and Deep Seek. Consider deep. It essentially kickstarted the AI momentum for China by shocking the world. But the Chinese government didn't put deepse on some pedestal. At best they invited the founder to some meeting [music] in which multiple entrepreneurs were present. Today they have multiple frontier companies competing globally. But in India the government selected one

2:20 company Servam under the India a mission before even achieving anything. To begin with Seram was founded in August 2023 by Vive Ragawan and Pratish Kumar. Vive was a key architect behind India's other infrastructure and Pratush has a background in artificial intelligence from IIT Madras. When they founded Serum, they marketed themselves as people who are building AI for the country. Their LinkedIn bio clearly says that they are on a mission to build the bedrock of sovereign AI for India. In April 2025, the Indian government selected Serum under the India mission to build India's sovereign foundational model. So overnight it was given national identity without needing to prove their capability or performance. So far they launched multiple a models

3:06 from 2023 to this year. They launched a model named open hati in December 2023. But the thing is it was built using metas opensource llama 2 and then they launched another model serum m in 2025. Again it was built using mistral the a model of a French company. After this model was launched, Indian origin Silicon Valley investor DDA publicly expressed the embarrassment on X. It just got 23 downloads in 2 days on hugging face. And this year in the prestigious India summit organized in February, they announced two foundational models trained completely from scratch. Serum 30B and Serum 105P. But here again, [music] engineers figured out that they used Deep Seek's open-source blueprint to build these

3:53 models. So they are fundamentally relying on foreign models to build their model. Now of course we can say they are building voice agents, texttospech models, document digitization models etc. But that simply means they are not competing globally at all. Which means India is not even in the air race. We have given up at least for now. But the problem is not just with artificial intelligence. The question is why India is so behind in building advanced technology. We don't have a single tech company on the scale of Google or Microsoft or at least Chinese tencent. Fundamentally, it's the country's economic, technological and educational problem. If we take the top AI companies, they are all either American or Chinese and they are the two largest economies in the world. We are not even close to China. Relatively, we are a

4:39 poor country. We don't have enough money to aggressively invest in technology. So, for India, lack of economic growth is one fundamental problem. The second problem is the lack of technological intent at the government level. In 2024, China has actually beaten the United States of America in R&D [music] spending. China is at $1.03 trillion while the United States is at $1.01 trillion while India spends just around 75 to 80 billion. China spends more than 2% of their GDP in R&D while India spends less than 1% of our GDP in R&D. America and China don't treat technology as something to be handled by private market forces alone. Silicon Valley is actually the byproduct of the government policies of the United States. The US

5:24 government was the main funer and the guaranteed customer of Silicon Valley in the early days. In China, the government actually sets the technological direction. Solar, EVs, batteries, highspeed rail, telecom gear, ship building, etc. And then they deploy cheap credit, land, local government subsidies and guaranteed domestic demand. Now in their latest 5-year plan, they have set the agenda for artificial intelligence. But in India, although the state created IIT, ISRO, Indian Institute of Science, etc., it failed to create tech company ecosystems through state policy. Even today the government is largely silent on this and they're more interested in supporting one or two legacy business families rather than young scientific talent and the next weakness is education. One common

6:11 argument is India doesn't have a talent problem. India has a lot of talent. In my opinion that is nonsense because the reality is we also say that majority of graduates are not getting jobs. So whatever the talent exists in the country, it's mostly because since India is such a big country, even if a tiny percent of students are talented, that seems like a lot of talent. But for the vast majority of students, India's education system is rotten to the core. In the QS world university rankings, there are 26 American universities, six Chinese universities, but there is not a single Indian institute or university in the top 100. Despite our population being the highest in the world, the top Indian institute is IIT Delhi ranking at 118. We have IITs but they are no match

6:58 to Stanford, Harvard, MIT, Wharton, Yale etc. So we are in severe short supply of elite educational institutions. As a result of all these even the few talented students leave India and become successful in US tech companies. So look at the whole chain. weak economy, low R&D spending, lack of government will to create a tech ecosystem, weak universities, brain dream and weak economy. Again, it's a vicious cycle and serum is just one visible symptom of that. Even the venture capitalist ecosystem in India is not big enough and ambitious enough and they want fast and safe returns. So, India needs an entire revamp. The overall economy has to grow much faster. We should raise our R&D spending. The government should become

7:43 proactive. It should support private tech companies aggressively and become the first customer. And then we should build as many world-class universities as possible which produce original research papers. The venture capitalists must become more aggressive and think long term. And also we should build in silence and let results make the noise instead of getting caught in nationalistic hype waves. So this is our take on the situation. What do you think? Comment below. [music] Heat. [music] Heat.

8:28 >> [music]