#334 Abhishek Singh: The $1.2 Billion Plan to Turn India Into an AI Superpower

0:00A couple of decades ago, India and China were roughly equivalent. Both were hungry, both were opening their economies, both were sending brilliant young scientists out into the world. Then China hit the accelerator. It built labs the size of small airports. It repatriated senior researchers. It handed them compute, and it treated AI like a national moonshot. India, meanwhile, rode a different wave. IT services and outsourcing, while its English language advantage made leaving easier than staying. The country produced world-class engineers, but it didn't build a world-class research ecosystem to hold them.

0:42That gap mattered. Without deep investment, labs, compute, mentors, long-term research culture, India became a training ground for the global tech industry rather than a proving ground for its own. Some of the brightest minds left and never looked back, going on to lead Google, Microsoft, and many other tech titans. This was never a talent question. But India is finally treating AI as infrastructure, building national data centers, national data repositories, and funding AI research in a bid to catch up with China. I'm doing a series of podcasts on that question. Why did India fall behind after starting shoulder-to-shoulder with China?

1:28What did decades of underinvestment cost them? What will it take to build labs that people return to, not escape from? And with the US and China so far ahead in research output, compute budgets, industrial adoption, can India become something more than a secondary player? In the first of these episodes, I sit down with the head of India AI Mission, Abhishek Singh to talk about India's national program to rapidly scale its artificial intelligence ecosystem, from building vast data centers and the AI-KOSH data platform to fostering homegrown talent and creating centers of excellence across the countries.

2:13If you're curious about how one of the world's largest and most diverse nations is tackling the opportunities and challenges of artificial intelligence, both at home and on the global stage, this conversation is not to be missed. Could you introduce yourself to listeners in the United States and alike the world about who you are and what your role is in the AI mission and what India's AI mission is? I'm a civil servant, a career civil servant with almost 30 years of experience of having worked in diverse roles across the country, various states, I worked in the extreme northeastern state of Nagaland, I worked in Uttar Pradesh, I worked in Government of India for long ten years.

2:59I primarily spent a lot of time in fact, in the field of technology. In fact, I may say that in almost every role that I have had, I've worked on projects which deal with using technology for improving governance. In my present role where I'm leading the AI mission, along with my additional role as additional secretary in the ministry and also director general national informatics center i'm responsible for implementation execution of india's 1.2 billion dollar ai mission and this is my fifth sixth year in this ministry in this tenure prior to this again i spent three years this means even i was leading uh digital india projects the common service centers project the other i was part of the adhar project has been involved with the conceptualization of the india stack implementation of various projects and then whether it was my girl whether it was dg locker the several of the at least 10 to 12 of india stack projects i have been directly involved with leading them i also led the karam yogi bharat i got training platform the technical platform that we built so primarily a civil servant with but a lot of experience of working in the technology sector within the government.

4:14Yeah. And how long has India AI mission existed? India AI mission was, of course, approved formally by the cabinet last year and by the government. The funding was approved last year in March, but we were working on conceptualizing India's strategy of AI. We had come out with this national strategy of AI in 2018. We were taking up projects in various dimensions, like for building compute, for building data sets, for building applications that was ongoing. But the big mega India AI mission was approved in March of 2024. And since then, I have been leading it. But on AI-oriented projects, I have been working for almost close to four years.

4:52I see. And you mentioned 2 billion. Is that US dollars? 1.2 billion US dollars. 1.2. That's the government spending. And in addition, the private sector is also investing in AI, like, for example, for enabling the compute infrastructure, while government is kind of providing support for the end users and government's commitment is as I mentioned across the seven pillars 1.2 billion dollars but the private sector in India has already invested around 2 billion dollars for enabling the compute that is available and in addition various government of India departments like health, education, agriculture and state governments they were also investing money so they would be anywhere between 2 to 3 billion dollars so total public investment uh on ai in india would be somewhere around five to six million dollars yeah i know there's seven pillars seven pillars it's a lot to talk about uh so i wanted to focus on uh on the data center build out i think right now you have three or four gigawatts of of data center capacity and the projections I've seen have reached 10 gigawatts by 2030.

6:06Is that purely Indian government or does that include, I know the Stargate, the US Stargate program has promised to build I think a billion dollars worth of compute data centers in India. So that building up to 10 gigawatts, does that include everything or is that Indian government only? It does include everything. It does include everything. And actually the installed data center capacity in India is presently the ones which is operational across the private sector, across the government is only one gigawatt. But there are a lot of committed investments which are there so we have already data center capacity which is being announced and which is currently being built that would take the overall capacity to 3 to 4 billion 3 to 4 gigawatts like for example microsoft has announced an initiative for building a three billion they announced three billion dollars for their gigawatts then google is looking at expanding the data center capacity in india geo announced a one gigawatt data center that they're building in jam nagar And then Stargate, we read about the announcement where part of the Stargate investment that they are building compute capacity globally, part of it will come to India.

7:27That will be again probably another one gigawatt. So we do believe that by 2030, we should be able to reach somewhere around five to seven gigawatts. I see, by 2030, five to seven. Okay. And then India has a lot of people. You have a lot of data. you have this amazing fingerprint database it's kind of the wonder of the world can you talk about I guess it's called the India kosh the the the data platform that that's pulling together data from breaking silos and creating a national repository see one of the key components for building ai is data yeah like of course the other two being talent and compute talent we have a plenty like we are known for to be the it garage for the whole world and compute is something that we are augmenting but most importantly the real ingredient or raw material for building ai models training models or building ai applications is data so while we have had a lot of digitization with the adhar project the identity project you spoke about the digital payments project and digital projects in almost every domain but very often the data set that we had was in many ways not AI usable.

8:48In order to make the data sets available and accessible in a form in which AI developers and application developers could use them, we are setting up this platform, the India data set platform called AI KOSH. KOSH in Hindi means repository. So in a way it's a repository of data sets and we are getting the public sector departments, whether it's agriculture or healthcare or education and all government in a department, state governments to share data on this platform. And in addition, we opened this platform even to the private sector entities. If there are telecom operators or other private sector players, hospitals, schools, if they want to share the data, they can also do that.

9:31And when they do the data sharing on the platform, the policies ensure that ultimately the data owner the person who is the ip over the data he has the rights he has a control over the mechanism of sharing of data whether they want it to be an open domain whether they want it to be restricted in case some costing is required or if it is restricted for access only to public users or to researchers so that entire control with regard to who has access to this data is provided by the by the is done by the data owner so this ensures that people have an equal right and then we also a lot of tools on the platform data anonymization tools privacy preservation tools so if anyone is publishing data sets on the platform he will have an option of cleaning the data and then putting it up on the platform so that you are compliant with the with the digital personal data protection act and other laws and in addition we also what we done is along with datasets we hosted models on the platform and provided a sandbox kind of environment so that if there are researchers or entrepreneurs who want to try out certain applications or build ai applications on the fly on the platform they can access compute they can access datasets they can access tools they can access models and then build ai applications on the on the platform the ai coach platform yeah and and for uh researchers i mean i'm particularly interested in research how do the IITs for example access the computers or I think I read there's a voucher system kind of like Google has credit system for academics so like compute was identified as one of the bottlenecks that we had in our AI story like in last year the number of GPUs that was available in India was around 500 GPUs only.

11:23That was a big, big constraint with regard to doing anything related to AI. That's like the fundamental building block that you need if you want to move forward. So what we did was that instead of government buying compute, what we did was incentivize private sector players to invest in the AI compute infrastructure, get the NVIDIA and other GPUs, the chips there, set up the cloud, set up the infra, provide it to the end users and we came up with a policy that the end users will be subsidized to the extent of 40 percent of the compute cost so that and through this tendering bidding process we got very attractive competitive rates so today almost 38 000 gpus have been made available through this mechanism to end users who have to pay only an average 65 rupees which is less than a dollar per GPU per hour compared to the international prices of$2.5 to$3 per GPU per hour.

12:18So that way researchers in IITs or even small engineering colleges or entrepreneurs or startups who are working on AI applications and models, they can access the compute, they can come to our portal, they can give the requirement and then the system will allocate them compute and they will be able to use that. They pay only 60 % of the cost which is much lesser and we bear the cost up to 40%. Yeah. And you mentioned talent. Certainly, India has a huge number of engineers, almost as many as China, I think. But in talking to the IITs, there is a historic problem that the best and the brightest go overseas for a PhD and don't come back.

13:08these Indians running the top tech companies in the United States. And so there's also a problem with Google coming in and Meta coming in and hiring away the top talent from the IITs. And so the concern is there isn't the human capital at the IITs to train the next generation of researchers because they keep on getting pulled away. I saw in my reading that there is a proposal, if not a program, to offer globally competitive salaries to some researchers. How do you address that brain drain out of the IITs? See, this is a very, very valid problem and challenge that we have. we do lose our best of the talents to the Western American universities and corporations.

14:11In fact, very often they give them such offers that it becomes very difficult to refuse. And if you look at whether it's Microsoft, led by Satya Nadella or Google, led by Sundar Pichai and hosts of other companies at various levels. Indian engineers have been part of almost every big tech firm in the United States and elsewhere. even in Europe, the companies which are there. We do have Indian talent going into the universities also for doing research. So we do have various ways and means to try to solve this problem. We have had extensive consultations with regard to many of the experts who have gone.

14:50Like what will it take for them to come back? One of course thing is that maybe funding. So through India AI mission and through the research development grant scheme which has been approved under which funding support is given to those Indian experts who are returning to take a projects here. So that is, that's, that's of course ongoing. And some of them are like making like one of the teams which got awarded for building foundation models of India are like Stanford graduates who are wanting to return and build something here. But the biggest challenge is that very often these researchers, they say money is just a small part.

15:25What they also need is a complete ecosystem of support which is available in the valley, in the Silicon Valley. So because there's a good network of venture capitalists and academic professors, experts. So that is something that we're currently trying to build up in India as well. We do have certain professors who have returned and who are leading such initiatives, whether there's Professor Ganesh in IIT Bombay, Professor Mitesh Kapra in IIT Madras, then some of the faculty members have incorporated companies so while reverse main drain has started to happen but what we need to do is to provide many more incentives especially expand our investment in R &D provide more financial support create a network of academic experts across institutions within India who can work together have other means of partnership with American universities where even if there are experts in the US, how they can collaborate with researchers here and work together on joint projects.

16:28Those are some things that both Ministry of Electronics and IT here and Department of Science and Technology are working on. And hopefully we will see in the days to come. And especially with the current geopolitical situation, there are a lot of uncertainty in the United States. So that's also in a way an opportunity for us to, the H-1B visas are restricted, then maybe the silver lining in the cloud will be that some of our talented fellows will not get their visa, not get their green card and they will come back and then they will build something of repute here and we will provide all the support that we've been required.

17:00Yeah. In this ecosystem you're talking about, I know one of the pillars is the centers of excellence. What does that mean? Is that building up AI schools within the IITs? Is that startup centers around the country? I mean, what is centers of excellence? The Center of Excellence is a program that we have started in partnership with the Department of Higher Education, which felt that while there's a lot of research which is going on in a particular sector in silos, like for example, AI in healthcare, maybe some professor is doing some work in one IIT, someone is doing some work in another IIT, and then there there is no connect between all of them.

17:48And maybe each one of them is solving a different part of the whole problem. So if we can bring them together in a hub and spoke model, so what happens is that one anchor institution becomes the nodal institution for that sector. So for example, the Center for Excellence in Agriculture, IIT Roper has been given that responsibility. So IIT Roper, the experts who are there at that IIT will form the nucleus. and along with that there will be other affiliate institutions which will work with the with the central institution and they will share expertise they will share data science they will share applications and together they were able to build something which is much more worthwhile for the whole sector similarly we have a Center for excellence of smart mobility that's incubated in IT we have a Center for excellence in education which is coming up at ID Madras then we have a Center for excellence in health care which is being jointly developed by AIMS the Holiday Institute of Medical Sciences and in Institute of Science Bangalore so that we different academic institutions they provide the leadership and bring together all the experts in a particular domain so they are able to work in a collaborative manner and add value to the research or the application that they're developing yeah but this education remains a bottleneck i mean there's a gap uh between a number of uh ai experts or senior researchers and the number that's needed is that right or see like uh like uh of course of course there is a so so what we've also done is that enhance the number of fellowships that we have like expanded number of seats that are there in the iit's the engineering colleges to provide more fellowships at undergrad level at the master's level and at the phd level invest more money in compute provide more support for getting data sets provide more support for getting expert advice from from leading scientists in the field so in a way in a way it's like a several steps all together which are being targeted to ensure that we are able to train our human resource in the domain of AI, data science, machine learning, deep learning, and be able to compete for the best.

20:11Yeah. And it seems, I mean, AI is moving so quickly that you can come in and very quickly be at the cutting edge. But why did India take so long to start this initiative? Whereas you obviously have been watching the United States and China invest heavily in compute and data and human resources. Was there a systemic reason or a political reason or why is India late getting on the bank rocket? See, we are, one can say that we are late, but as I mentioned that we had come out with our National AI strategy in 2018. Yeah. And then of course, there was a disruption for some time because of COVID and other reasons.

21:11And that required allocating resources to sectors that needed the most. Yeah. So, we were working on it plus at the same time finalizing the plan. But what we have achieved in the last one year of the approval of the program is quite significant. And we do hope that we will be able to not only catch up with the United States and China, but maybe leapfrog, especially with the expertise that we have and with the efforts that our researchers are making. Yeah. So the program, the AI, India AI mission, what's the timeline that you're looking at? See, the whole program that we have is for five years to begin with.

21:52But and we are already in the second year. So we do hope that by early next year, when we are hosting the Impact Summit in February 2026, we'll have something very tangible to show to the world. And the LLMs that our startups are building will be different from what we have known so far. And it will be customized for Indian requirements. It will be built for Indian needs. it will be more language focused it will be more voice enabled so that's the way our researchers are working towards building something so we do have faith in their expertise in their abilities and their talent and we are hopeful that what they build will not be a copy of a chat GPT or other tools that we are used to but it's something which will be which will be truly designed to meet the challenges that we have in India yeah and they're unique challenges because of the diversity of language and diversity of religion.

22:52I mean, there's been a lot of talk of sovereign AI of different countries building their AIs for their domestic cultures. Is there a government effort to build sovereign LLMs or AI systems, or is it happening in the private sector? It's happening in the private sector and in our academic institutions, but it will be funded by the government. We are providing financial support and here when we are providing funding support, we are providing the entire compute costs. So we already have four efforts which are currently ongoing. Eight more have been finalized which will likely to be announced very soon.

23:34And one of them is actually will be at one of the IITs. So it will not be a startup driven effort, but it will be an academic institution driven effort. and since the IIT which is technically a government institution that we are trying to position as a sovereign effort for building an LLM and a lot of funding is involved there and lot of expertise and a lot of experts from across the country are coming together to build such an LLM. Yeah and can I ask which IIT? Let that prerogative with the minister when he will make the announcement. Okay and the event next year I'm hoping I'll be invited Yes, we love to have you at the Impact Summit in 19th and 20th of February because we co-chaired the Action Summit which France hosted this year which was in the series of the Bletchley Park and the Seoul Summit and the Paris Action Summit.

24:28the next summit which we are calling it the AI Impact Summit which will look at the impact of AI on the key sectors on people, on planet, on progress so that will be deliberated and we are expecting experts from across the world, from industry, from academia, from government to come in so we would love to have somebody like you to join in and you will get many more opportunities to talk to many more experts from across the world Yeah, China you know is the the big behemoth in all of this. I mean, certainly the United States is. But China has been making moves recently to develop a governance organization that would include the quote-unquote global south.

25:18Would India participate in that? The way it looks is the U.S. is working through allies and also with kind of a carrot and stick approach to get countries to adopt their AI government structures and standards. China now is moving to create a sphere of influence of its own and it looks to me that it's it's going to be two two zones there will be the Chinese dominated zone in the US dominated zone where does India stand in that see like this is like a topic on which we have a lot of discussions debates going on for a long time and the way I see it converging, especially on something which China is also aligned and on other parts of the world is also aligned as maybe something that's driven under the aegis of the United Nations.

26:26Because AI is a global technology, there are stakeholders involved. We cannot exclude any major player from any governance framework because if that you do, then it will be limited governance. So the UN Global Digital Compact, the efforts that they have made, the UN high-level advisory body on AI which was constituted, they are coming out with an institutional framework which will allow all countries to have a stake in deciding how the AI governance will shape up. So we are more aligned to something that will ultimately work and our way to approach this subject has always been Indian way rather than American way and China and the Chinese way.

27:07So we have our own approach which works towards having a regulatory framework which ensures promoting, building innovative solutions that can solve societal challenges in various key sectors like healthcare, agriculture, education. and which also gives the Global South a voice in trying to ensure that the AI applications models are more inclusive, they are not biased, they incorporate the data sets of the large majority, including India, the country, the Global South. How do we bring them there? How do we ensure democratization of AI resources? How do we ensure that researchers from across the world get access to compute, to get access to data sets, get access to algorithms.

27:58So that has been our principle. And we have been, even when we led the global partnership on AI, we came out with a new inclusive partnership with the OECD and other countries. We do hope that ultimately that will be the approach the world will take. In fact, the discussions at the summits also have moved on from focusing entirely on safety and security to India's approach of having a balance between innovation and regulation. and if you look at even the evolution of the summit at Bletchley Park, it was only safety and security. At Korea, it became a sole summit. It became safety and innovation.

28:34And in Paris, it was about action and it included safety and security. It was there on vertical. There are other four themes also, which looked into public interest AI, future of work, sustainability. So those issues also came into the forefront. So what we have been doing from the last several years is that championing the cause of ensuring that AI is more democratized. AI addresses key social challenges, focusing on building applications that can benefit the larger humanity, working on tools and methodologies to restrict user harm. And that's the principle that we need to take forward. If these are incorporated in the U.S.

29:11White House action plan, to some extent they have already been. The White House action plan is also moved away from the AI diffusion rules which have come in January. And China is also like look at the Chinese action plan. It also reflects many of the things that we have been advocating at various forums. So we are hopeful that the global order that will emerge will be more inclusive and will reflect the views of India and the country. But India will be operating primarily through the UN framework, not not through the US.

29:48organizations or through China's organizations or will India take the lead and create its own governance and standards organization and to try and pull in the global south I mean India's perfectly positioned for that yeah we see the countries of global south are looking at India for providing a lot of support and we are committed to provide them support when it comes to like the dpi story is known for everyone like we implemented digital public infrastructure at population scale and at the g20 it was recognized as a strategy and one of the key declarations of g20 was that india went along with the g20 countries will build up a global dpi repository and that work is ongoing so here also in ai many countries of global south are looking at at leveraging india's expertise in building a global ai use case repository so we are working towards that the gpa we have been advocating the same story we have been part of the g7 friends group we have been part of the un high level advisory committee on ai of course we did not attend the china summit but the overall approach that we have been advocating is in many ways captured in the document that china has so i think we see a lot of alignment globally on the approach that india has been advocating whether it will be a un body or whether it will be a separate organization or a technical standards organization that is something that's currently being debated at various forums yeah uh and with the the recent warming of relations with india uh prime minister modi was was just there as we all know uh is is there any increase in cooperation on ai research or ai infrastructure uh between india and china the visit has just concluded so we will have to see what are the domains in which we will be moving forward we just had a visit of the prime minister Singapore today and that declaration does mention AI exclusively or working together creating joint labs and projects in this we have active partnership with the country like Japan with France with UK with the United States so currently that's the domain in which we are present people.

32:01Yeah, and then I'll finish up, but the 1.2 billion, how did you arrive at that number? Did you want more? I mean, everyone always wants more, but the government, does it have the resources to provide the funding that you need? And after the five years, is what are the funding goals? See we have a commitment to our honorable prime minister that whatever we need will be given to us. There's no restraint, there's no constraint of funding at today's time and age. So we had estimated that this is what we need basic minimum to trigger the whole ecosystem and in case we suppose this is the 1.2 billion we have got for five years in case we are able to exhaust the funding in say three years we can always go back to the government and get more funding so that assurance that we have that support we have of honorable prime minister he has always been egging us on in working towards building something that is state of the art that's like global that can that can be offered to countries the global south if there's an ai solution which can help take to diagnose tuberculosis or an ai solution that can help meet the needs of farmers or transform education, then we should build it and we should offer it to the whole world for usage.

33:31Yeah. And then my last question, you said that the goal is to catch up with U.S. and China. Do you have a ballpark for when you think that's possible? Is it five years, 10 years, 20 years? because China and the US are advancing at the same time the number of gigawatts data centers and all of that. I would not like to put a timeline to this, but the aspiration level is there and we are working towards that goal step by step, one step at a time. And as long as we are able to provide the enabling ecosystem, the role of the government is to provide the enabling ecosystem and we have faith and confidence in our researchers and startup entrepreneurs that they will be able to help us move forward.

34:19So we have done that. We launched a lunar probe at a fraction of the cost of what it costs. So we have seen achievements in several sectors. Sometimes it does take time, but we reach there. So the commitment of the government is there. So instead of putting a timeline, I will say that as long as we are firmly on the path towards achieving what we want to achieve, maybe we should be able to catch it soon.