I want to start with a core technical question. How is your leather jacket holding up in the Mumbai heat?
(Laughs) It’s fine. Thank you! I enjoy the heat.How prepared should India be for disruption in the IT sector, in terms of low-level coding being very rapidly getting disrupted by AI?
Well, the future of programming is going to be very different. We will do less programming of the computer and do more programming of the AI. And the idea of programming explicitly, telling the computer exactly what we want, is going to be less. And then letting the computer know what we intend and what we’d like to achieve will be more popular. And so we’ll do less coding and do more prompting.
This is an extraordinary time for India for several reasons. India has one of the world’s largest digital ecosystems and an enormous population of computer-literate engineers. This is just an extraordinary opportunity for India to reinvent itself from a back-office, IT cost-reduction industry to a front-office, AI-driven innovation ecosystem. And I was so inspired by all the companies that I met while I was here. The energy around artificial intelligence and the literacy and the capability that has been developed here in the last year (is great). In just the last year alone, Infosys, Wipro, TCS and many more that I met, the work that they’re doing in agentic AI, at the leading edge of artificial intelligence is really quite extraordinary.
There is a fear that AI will replace jobs…
Discover the stories of your interest
You know, AI is not going to replace humans anytime soon. There’s not one piece of artificial intelligence that can replace 100% of someone. But, it is the case that everyone can take advantage of artificial intelligence to elevate our capabilities. And so think of it as a tool that elevates our abilities to make us all superhuman. And I think that that understanding pervades in India today. Everybody that I see, the number of startups here, the understanding of the opportunity of artificial intelligence, the energy that’s here is really quite extraordinary. And so I’m excited about the opportunity for India. A third of Nvidia is here in India. Somebody mentioned that most of the letters of our company spell India. And in a lot of ways, Nvidia’s senior leadership is Indian. A third of our engineers are here. We’ve been in Bangalore for almost 25 years. India designs Nvidia’s chips, writes Nvidia software, designs Nvidia systems, and develops a lot of Nvidia’s algorithms.
The aspiration, the vision, the ambition to elevate yourself from an IT cost reduction industry, from a labour outsourcing industry, to become an artificial intelligence production industry, I think you’ve got to pursue that with all of your might. You don’t want to be a cost-reduction industry, you want to be a revenue-growth industry. You want to be an industry where, when you’re sleeping, you’re still making money. And it’s hard to do that when your business model is by an hour. And I’m excited by the fact that the leadership here recognises this extraordinary opportunity.
There’s no reason that the fundamental intelligence of this country is not encoded in intelligence that can be manufactured at scale. Every aspect of artificial intelligence, the natural resources are here. The digital economy is here, so you have lots and lots of data. You have a deep understanding of computer science, you have a deep understanding of computing. You have massive resources. And in order to be in the artificial intelligence industry, to manufacture intelligence, you need energy, you need data, and you need computer science expertise. All three exist right here.
And I remember fondly the first time I spoke to Modiji about artificial intelligence. And he had a great deal of curiosity about what it is that I did for a living. And I spoke to him about AI. And at the end of it, he said something that was profoundly wise. He said that India should not export flour and import bread. The flour is the natural resource of your country, which is data. The data of India belongs to India. It’s your natural resource.
The data of India encodes the knowledge of your people. Of course, it’s the language and the history, but it also encodes your knowledge and your culture. It belongs to you. There’s no reason to let anyone else harvest that, process that, and turn it into something of value called intelligence, digital intelligence. You can do it yourself. He also said that India should own its own AI cloud — meaning that the AI infrastructure, the processing infrastructure, the factories — we call them AI factories. The AI factories are part of your national infrastructure. Just like energy, just as roads, and communications, all of those networks should be built here. And it should be part of the national infrastructure. And his understanding of it was quite inspiring to me, quite impressive to me. And when I come to India, I have had the opportunity to update him several times on the progress of artificial intelligence. And you can tell that he is honed in on this idea that AI can lift the nation. And the example that he spoke about was agriculture. The vast majority of the country is in the industry of agriculture. And if we could put AI in the hands of farmers, so that the use of AI enhances crop yields and make better predictions of the weather and the yield of crops, that could lift the productivity of the people. So I think in a lot of ways, whether it’s the national will, the inherent capability of this industry, the IT industry, and also the digital economy that we’ve already created over so many years, the ingredients are right, the ingredients are here for you to take advantage of this AI revolution. (India should) manufacture AI and lead the AI revolution.
You are bullish on applications such as robotics and autonomous driving. Now that you have experienced India’s traffic, do you think Nvidia can make chips powerful enough to handle autonomous driving in India?
Well if somebody can conquer autonomous driving in India, you can be assured that autonomous driving will be solved everywhere else in the world. Of course, we also know that autonomous vehicles use the wrong sensors for India (jokingly). Cameras are obviously insufficient. You can’t use computer vision, you need sonar. Because, you know, here in India, you have to honk. You drive by sound, not by sight. You can close your eyes and drive in India (laughs).
You started investing in the CUDA (Compute Unified Device Architecture) platform back in 2007, and for years afterwards, Wall Street analysts kept asking you where’s the ROI in this. What gave you the conviction to continue investing in the platform?
At the core of the company, we believed that a general-purpose processor can’t be good at everything. And we were in the beginning of a new technology revolution like the microprocessor. The rate of progress was so great that the general-purpose CPU wasn’t good enough. We felt that there are many problems in computing where the general-purpose sequential processing capability of the CPU wasn’t the right fit. And there are many algorithms that could be offloaded from the CPU, the sequential processor, to be accelerated using a parallel processor. Now, we believed that in the very beginning, and we believe that today. And if you believe in something, unless some fundamental new insight were to change your belief, there’s no reason to change your belief. And so we’ve pursued it for about 33 years now. And finally, our day has come.
We invented accelerated computing. It sits next to the CPU. It doesn’t replace the CPU. The things that are sequential continue to run on the CPU, but the things that could be accelerated in parallel acceleration, parallel processing, could be offloaded. It has taken us three decades because you can’t just put a GPU in a system and all of a sudden everything gets faster. You have to actually change the software applications on top. It is the reason why so many of our endeavours require a brand-new software stack. Nvidia is in a lot of ways, really a computer algorithm company.
We’ve reduced the marginal cost of computing in the course of the last 10 years by 100,000 times, to the point where researchers said to themselves, why don’t we take all of the data in the world, take all of human knowledge, take all of the entire corpus of human knowledge, and give it to the computer, and let the computer figure out where the knowledge is, what is the knowledge that’s represented inside, by learning the patterns and relationship of every single piece of information that it was able to learn, basically, machine learning, or what we now call large-language models.
It’s easy to forget now because of Nvidia’s stupendous success at this point that how hopeless a startup it was. You used to start your meetings by saying that our company is 30 days from going out of business. From that to today where concerns are being raised that Nvidia employees have become so wealthy that they are being criticised for being a bit complacent by some… Having seen this journey, what is the greatest motivator—fear or ESOPs?
I woke up this morning thinking the company would be 30 days from going out of business. And that hasn’t changed. No one in technology should ever feel too comfortable. Technology changes incredibly quickly as you know. And artificial intelligence is the single largest industry that the world has ever known. And the reason for that is because intelligence is the largest industry the world has ever known. And so it’s natural that we have a lot of competitors. And so we have to make sure that we don’t take our position for granted. And of course, we grew everything from nothing. And so I know what it feels like to have nothing and to be nowhere. And those feelings don’t leave you. I grew up poor and our company grew up poor. And I’m sure that there are many people in the audience who started from rather modest beginnings. When you start with modest beginnings, those feelings never leave you. I don’t know about you, but I still enjoy leftovers. And so I think you want to stay modest as a person. You want to stay modest as a company.
The second part is that you might be surprised that many of the employees have been rich for a very long time. I’m looking at a few employees right in front of me right now. They’ve been wealthy for a very long time. And yet they’ve worked incredibly hard that entire time. I’m not working for the money. I’ll be honest. I’ve been wealthy for a very long time. And yet I work harder than ever.
Let’s talk about energy. AI needs a lot of it. What are the sustainability practices that Nvidia is adopting and where are we headed?
First, we need to improve the energy efficiency of our computing as fast as possible. The second thing is to realise that AI doesn’t care where it goes to school. AI doesn’t really need to be close to us. And so we can put the AI data centres near where we have excess energy.
You know that the world has more energy than we use. Obviously, a lot more energy comes from the sun than is actually used. But we don’t have excess power because people like to live in certain areas. And so we can move the data centres, build the data centres farther away.
And then the third concept is to remember the goal of AI is not to train the models, which consumes a lot of energy. The goal of AI is to apply the models to be more efficient. Remember, I’ll give you one example. We can predict weather 1,000 times, 10,000 times more energy efficiently than using computers and simulators.
Net-net, however, I will tell you that the world will be a lot more productive in the future. We will be much, much smarter about using energy. However, my hope is that the amount of energy that we use for artificial intelligence increases as a percentage of total energy over time. And the reason for that is very simple. I hope that the production of intelligence is a very large industry. Because we all hope that we produce intelligence. The goal is not to produce cement. The goal is not to produce steel. The goal is to produce cement more smartly, and present a new type of material that could replace steel. Our goal is to be much more smart about agriculture so that we can do everything a lot more efficiently. And so my hope is that if we end up using more energy for intelligence, less for cement, steel, less energy for sitting in traffic, less energy for all those things that consume energy that we’re using.
What happens after we achieve AGI (Artificial General Intelligence)? And are you concerned about effective regulation around the development of AGI?
We should regulate AI. We should regulate AI in the context of every application. When you use artificial intelligence as an accountant, that accountant should be regulated. Similarly, for lawyers, doctors, etc. AI should be regulated in the context of its use.
Your first question has to do with artificial general intelligence. My hope is that we will all be surrounded by super-AIs, super-intelligent people. And what happens when you’re surrounded by super-intelligent AIs? Well, that’s exactly my thing. Remember, I’m surrounded by people who are far better at what they do than I am. I’m surrounded by super-intelligent people, and yet I have no trouble working with all of them. And in fact, aren’t you supposed to surround yourself with smart people? Smarter people than yourself? And why wouldn’t you want to have assistants that are super-intelligent at the skills that they do?
Can AI help societies like India to solve fundamental challenges of poverty and development? Should countries be starting to think about universal basic income?
As you know, the vast majority of Indians do not know how to program a computer. The ability to program a computer is one of the greatest economic capabilities today. Most of us in the computer industry have access to a capability that most people don’t have. And so, on the one hand, very few people can program a computer, but everyone can program an AI.
What I’m saying is, for the very first time in history, the computer is an instrument for everybody. Not just for people who are privileged or somehow were better educated or learned how to do that. And so I think the technology divide is likely to be reduced, likely to be eliminated. But it’s more likely that AI will elevate the capabilities of everyone than the top 1%. Most people thought initially that AI was going to threaten blue-collar jobs. But it turns out it is likely to be much more challenging for the white collar. And the reason for that is because everyone is now elevated.
What have been your biggest takeaways from this trip to India?
Back office to front office. An industry of cost reduction to an industry of innovation. An industry of labour to an industry of invention and deep tech. An industry that can only make money when you spend hours, to an industry that can make money when you’re sleeping. That’s the industry that I’m certain India will become.
I came not expecting anything, but I leave with incredible enthusiasm and optimism that across the entire IT industry here, all of the technology companies here, the CEOs that I met here, the determination to reinvent the IT industry, to take advantage of this generational opportunity, the once-in-a-lifetime opportunity to reinvent, to take advantage of that, and translate it to take advantage of the natural resource of this country, of this nation, and propel it forward. I am super-energised by it.
