First of all, I want to confess to you that I am not an AI expert. I know nothing about AI. I am basically a person, an entrepreneur who runs businesses. But I have a passion for technology. I have a passion for innovation. And that has been the basis of our business and business transformation over the last 50 odd years. I have actually been working in this company. This is my 53rd year. So it's been a long, long innings. So I have seen very, very large transformations in the last five odd decades that's taking place.
I had an opportunity to address this forum in 2020, I think, and we were in the middle of the pandemic and a global reset at that time. I shared my thoughts on the concept of what Atma Bharat is. That was the theme that the honorable Prime Minister had put to the whole industry and the whole society and what should be the priorities for Indian manufacturing. My space is only manufacturing. So I will only talk about manufacturing and then give you some examples of what we are doing in terms of how we are using some amount of, I wouldn't say AI, because when I heard Mr. Rao or Dr. Rao talk about this just before me, he himself said that he doesn't know what AI means, but it's a synonym that everybody's used now very loosely. So I will just say AI.
Five years ago when I made this talk, I said you need two, three things in Indian manufacturing to catch up with the rest of the world. You need speed, speed, and more speed to do things. You need that because if you want to remain relevant and in sync with a rapidly evolving digital and AI technology landscape, more so in the backdrop of the nationalistic sentiments that were running at that time. So today we are in 2026, almost five years later. That speed is being powered by what is loosely referred to as AI. But honestly, I don't see much of AI in the practical world anywhere except what you see in terms of ChatGPT or some of the LLMs that are there by Google, by Grok, by Meta, and other people.
But let's understand, as somebody who looks at business and wants to bring in new technology and innovation in business, it's inevitable because if you don't do it, you're going to be extinct like the dinosaurs. So one has to work hard towards maintaining one's own levels. Today, as a business, we are in the business of metal forgings for automotive, aerospace, power, oil and gas, shipbuilding, and we are the largest forging company on this planet. We are number one in this business today worldwide. And we have remained that way because in the old days, when I started working in the '70s, we were working with steam. Steam was the main power source for our forging hammers.
Then came electricity, because steam was very expensive. You have to run boilers, you have to burn fuel. It was expensive and it was bad in terms of environment and the ecological output that came out of it. And then came automation. Till 1988, we had no automation, we had no robotics, no nothing in our company. Then came robotics, then came automation, and then came Industry 4.0 in the last five, six years. And now we are on the cusp of this new technology, which is a combination of digital and using algorithms to solve your problems, what we would call loosely as AI.
You know, yesterday I spent a couple of hours watching on live TV the discussion that was going on in Davos in Switzerland. I've been going to Davos for 35 years. My last trip was in 2020. After that, I said enough is enough. But 35 years, and Mr. Morty used to be there in Davos for a long period of time. It was fun those days because we had fewer people, had better interactions with the universities, whether it was MIT, Stanford, or people from Harvard, technical people. You could meet customers in one place that otherwise took you six months to get an appointment. So it was a fantastic place.
Today it has become a jungle, and if you saw it yesterday, it has now become a place where one nation, I don't want to name this, is demonstrating its economy on everybody else through everything that it has got. It was a pretty bad demonstration yesterday, in my personal opinion. But hearing what people were saying there, there are two or three issues in my mind as somebody who wants to use the products of AI in manufacturing. We are a user. We are not involved in developing, or we're not involved in anything else. We want to use it for two, three simple reasons.
Number one, we think that it'll improve our productivity. But this is, from what I heard yesterday, a big question mark because nobody has yet proven that this can improve productivity. Yes, there are a lot of use cases, but no use case that is defined where productivity in manufacturing, especially, will improve with AI. So the question is, will it happen, will it not happen, because a lot of big investments are required even for a manufacturing company to get into AI.
We have set up an AI incubation center. Azog has seen this in Pune. We have 100 odd engineers. I made one simple rule there: nobody above the age of 35 can work in that place. Only for young people. And as we heard Professor Rao talk about, seven-year-old youngsters are able to code Python and solve problems. And we find that people who are in their 20s, late 20s, early 30s, are far more efficient in looking at the AI solutions that we want to implement in our business. So this is one problem, whether the productivity will happen or not in manufacturing.
The second is AI requires a lot of other infrastructure for it to succeed. We are seeing huge investments in data centers today all over the world. Mr. President Trump yesterday gave an example of Meta's data center that they are building in the US. He said which is the size of a Manhattan island, and spending close to $500 billion on that. The point is, the investment in data centers, nobody knows when the payback will happen, and that payback is dependent purely on the life cycle of the chip that they use. Already in India, we are seeing that people who have built data centers using the H100 chip are losing business very rapidly because there are faster chips available which are coming in in the new data centers.
So this is the second problem that we have: that we don't know what the investment ROI is going to be and who's then going to fund this investment, because you need all this backbone to be able to apply artificial intelligence in one's business. And the third is energy.
Now, as far as energy is concerned, if you look at it from a conventional point of view, you would need some enormous amount of additional energy in India. And it's therefore necessary for us to look at the nuclear options which have been approved by the government. But the problem is that if I put an application that I want to set up a 100-megawatt nuclear plant, for which technologies are now available in the US for SMRs, for example, it will take five years for the government to approve my application. Mr. Trump is approving this application in 15 days to two months at best.
So we need to get our regulatory system on board to be able to meet with the requirements that this new technology proposes. So this is not a technology that is a transient technology. This is a technology that is going to rupture, if I use Prime Minister Mark Carney's language, the speech he gave in Davos. I've never heard a better speech in my life than what Mark Carney gave yesterday in Davos. I've heard many speeches in Davos. I've seen Yasser Arafat and the Israeli Prime Minister Bengurion that time meet together and try and solve their problem, because it never got solved. Ultimately got about 70,000 Palestinians killed in the last one and a half years, and still the problem is going on.
But we have seen all these big events in Davos, people talking, making progress, making speeches. But what Mark Carney said, fundamentally two things: don't believe in institutions that you have been used to, because what we are going through is not a transient from one level to another level. What we are going through is a rupture. The same thing is with technology. When you look at AI, it's not a transition from some Industry 4 to digital and then to AI, whatever that AI means, I don't even know what it means right now. It's a rupture. And the problem with this rupture is the hegemony of two countries or two states, United States and China, are going to drive all this.
In India, I heard Minister Vishna talk about this yesterday in Davos. He's talked about putting all the five layers that are required for AI transition in India. And I'm happy that they have taken a decision to follow a small language model instead of a large language model. They believe that smaller solutions in different areas will work better for us with far less cost than large language models. You've already seen large language models fail, like the Llama. It has failed completely, which came out in April.
We heard Professor Ud's lecture a little while ago, and he talked about how AI can give you erroneous results if you work in a particular direction. It's very simple. It gives erroneous results because it is based on the data that it has got today, which is only sequential. It is not based on the laws of physics. It's not based on the laws that the world follows. Now that is going to be the new AI that will come out whenever it comes out, and that will change the whole direction of the world as we see it, and applications for people like us in the manufacturing sector that we are in. Google is working on it. That's what they have announced, but we have to wait and see how it comes.
So saying this, I will just tell you a couple of examples of what we are doing with AI, how we are using this in our manufacturing sector.
So this is the most important example of what we are doing. We have said, how can we make our factory completely autonomous? How can we have robots talk to the machine, the machine talking back to the robots, and robots controlling a set of machines without human intervention? So this is something that we are working on, and this requires a lot of compute power. So we are using Nvidia compute power. It requires smart robotics to integrate with normal robotics, because just normal robotics can only do programmed functions of pick, place, or whatever else that you need to do. It cannot think about what needs to be done when something goes wrong.
So this is one of the things we are doing, and we are hoping that by the end of this year we will have this running on our shop floor in terms of using some amount of artificial intelligence, a lot of smart robotics in putting solutions where manufacturing will become autonomous.
The reason for doing this is not just to reduce manpower. This will increase productivity. Our guess is it increases productivity by 30 to 40%. It increases quality. So we will not have things that go wrong with human intervention sometimes, and it just makes it a far better manufacturing environment. So we've been working on this. We have now got to a stage where our robots, our smart robotics, are able to control the robotics that are on the machine. The robots are able to talk to the machine. That means the machine controller is able to talk back to the robots. So any intervention you need to do, you don't need a person. It gets done automatically. We just have to put a couple of algorithms in place, and that's how use of AI will happen. And ultimately, this will lead to a dark factory in our system.
So this is one example. The second example, we are now using VR, AR, and AI in designing products, complete products. So once we have a 3D CAD design, a 3D model, just a simple CAD design, what every engineer does when he wants to design a product with the software, we have got, these are not our software, these are software that you get from outside, it gets converted into a 3D model completely, and that 3D model can then be blown up into components, and each component can then have an automated manufacturing sequence generated based on data that we have given to our AI agent in terms of how to produce it.
So our experience is that by doing this, we are able to make a brand new product, and this is largely we are using this in our defense business, make a brand new product in 30 to 40% of the time that it requires otherwise, with much less manpower. So we are building, to give you an example, this is an artillery gun. We are building a light tank. We are building an FICV for the Indian army. We are building an FIC that is tracked and that is wheeled. We are building turrets with guns, and these are highly automated systems. They are not simple systems. They are complex and automated, and therefore you need simulation and you need those kind of things. Thank you, Mr. Morty, to understand this.
And it becomes very easy once you have it in three dimensions for even an engineer to understand how to build this. So that's another example where we are using a combination of digital, AR, VR, and AI technology to create a production process for each component.
Next, this is an interesting thing. We work with the Indian Navy for underwater systems. Underwater domain awareness is one of the most important focus for our defense organization when they are developing various different products operating at depths from 200 meters to about 1,000 meters, that's roughly about three and a half thousand feet. Capture oceanographic and acoustic data for end users of customers. This data that we capture by sensors is raw data which needs to be then analyzed and processed to decision-grade insights.
Towards that, we are now building an AI-driven platform that will convert passive raw hydrophone data into real-time actionable intelligence for better decision-making. This platform helps in enabling persistent multi-sensor surveillance, improving source localization accuracy, minimizing operator cognitive load through visualization reports, and providing AI-driven classification for rapid threat categorization.
I mean, just imagine, we have a large coastline. Imagine that you've got enemy submarines, I won't say enemy, you've got submarines running around that coastline which are not yours, somebody else's. You also have big fishes that run in our oceans. Now, how do you recognize what is what? How do you recognize what is a threat? How do you recognize what to do when you recognize a threat? So those are the kind of things that we are developing here.
We have now got almost half of these things deployed already with the Navy in the Indian Ocean, in the Bay of Bengal. The next step will be to do it in the Arabian Sea. But this is now going to be required in a very large number because of, with all these trade wars and everything that is going on, the change in the whole geopolitical situation, the sea lanes are going to be very, very important for shipping routes and sea lanes. And for that, we don't have any protection today. And this is the kind of protection, this requires a lot of AI-based, it's pretty much like what you do in health.
If you do image sensing for health and then decide what is what, it's similar to that in health, except that you're image sensing the sea, underwater, and figuring out what is there and what is a foe and what is a friend. So these are some of the things that we are working on.
So we have a lot of work ahead of us. No doubt, we believe that innovation in creating new processes, we are not creating new products. We create new products by doing other things. But creating new processes which make it more efficient, which make it more reliable, which make better quality, is the way we would use artificial intelligence along with digital technology and along with the Industry 4.0 processes that we have. And I believe for the country going forward, this is the future.
If I look back for the last 50 odd years, as I explained, I've seen at least three or four major technological changes, but they were all transient changes. But this is a rupture. This is not transient. This is going to be learning everything new from scratch for many of us. And I hope that as a country we will be able to make this happen. It's important for India to be in this game. It's important for Indian industry to be in this game. And I think it's important for educational institutions like IIT to collaborate with industry together to make this happen. Thank you very much.
See, in the kind of manufacturing process that we have, largely, let's say, automotive components, critical components, aerospace components, I make turbine blades for Rolls-Royce out of titanium, these are rotating components that rotate at very high speed. We make the turbine shaft, we make crankshafts for cars, trucks, big engines. Now these are parts that have very, very close tolerances. And when you look at an aircraft component, for example, you can't even have, you have zero rejection. It's like zero tolerance for terrorism. You cannot have a 1% or half a percent or 0.1% rejection. You just simply can't have it. So you have to have a process first that is digital. Because without a digital process, if you've got humans intervening in every step of the process, you're going to have problems because people make mistakes.
I'll give you our example. In the '80s, we got our first export order to export a truck component to Japan, to Mitsubishi Motors. We were then using our old technology of hammers. So we made fantastic samples. The Japanese were very impressed. But when we made the first thousand pieces and sent, they rejected the whole lot and sent it back because we had metallurgical problems. When we investigated that, we found that when they take a tea break, like we had one coffee break a little while ago, instead of 15 minutes, they go away for 30 minutes, 45 minutes, and the material is sitting in the furnace getting overheated and the whole metallurgical property goes for a six.
Now this is the problem that we have. So we had to automate. That's when I made a decision that we will automate our plant. And today we have an automated plant. We don't have these kind of problems. But we now need to get over the problems of machine malfunctions. Now when a machine malfunctions, these are all machines with five axis, six axis. We have some machines with 17 axis working simultaneously. If something goes wrong, for a maintenance guy to figure that out takes hours.
If you have an intelligent system that is based on AI, getting data based on past data, what has happened, you can do this in minutes, and maybe you can also put in some program or algorithms or whatever you want to say to self-correct this problem. So these are the kind of things that we are working on. We put a simple thing in place five, six years ago when we started with Industry 4.0. We said we should have zero, in a factory like ours there are three kinds of problems on machines. One is problems with hydraulic systems, oil leakage, pressure drops, things like that.
Problem with electronic systems, something going wrong, some system busting off, and problem with the electrical system. And the last problem which you can't do much about is the wear and tear. But the wear and tear is predictable. So you know when to change what? But these three problems, we said we need zero hydraulic maintenance. We need zero electronic maintenance and zero electrical maintenance. And I'll tell you, almost in 90% of our shop floors we have been able to achieve that by using just Industry 4.0 sensors, sensors giving data well in time, people taking action on that. So this is the advantage of that and that improves productivity quite a lot. This is not rocket science, ladies and gentlemen, believe me, but it's a tool that we have in our hand to use.
