Cheaper AI compute will bring more work for TCS in building intelligence, digital twins, says Reguraman Ayyaswamy

Reguraman Ayyaswamy, Senior Vice President and Global Head of the Internet of Things (IoT) and Digital Engineering, TCS

Global chipmaker Nvidia is eyeing Indian IT services companies for long-term partnerships and banking on delivering its software and AI infrastructure to the end customers.

At the recent Nvidia AI Summit in Mumbai, founder and CEO Jensen Huang said, “India was earlier focused on IT, the back office, and delivering of software. The next generation of IT is going to be about producing and delivering AI.”

The country’s largest software exporter, Tata Consultancy Services (TCS), quickly became one of the early beneficiaries of Huang’s bullishness. TCS has unveiled a separate business unit with Nvidia to co-develop industry use cases on artificial intelligence (AI) with the chipmaker.

In an interview with Moneycontrol, Reguraman Ayyaswamy, Senior Vice President and Global Head of Internet of Things (IoT) and Digital Engineering, TCS, discussed a gamut of issues from the new unit to collaboration strategies with Nvidia, plans to purchase the upcoming Blackwell GPU (graphics processing unit) and the cost of generative AI development and more.

In terms of Internet of Things (IoT) and digital engineering, there are opportunities in a number of areas and to use generative AI as well.

Nvidia for now is a tool set. I will give a couple of examples. Take the software-defined vehicle (SDV) in the autonomous car industry. The vehicle's software development and a lot of ADAS (Advanced Driver Assistance Systems) features like autonomous car features, will get improved with generative AI. Especially if you have thousands of miles of video data and you have to train your algorithm on using the car in different conditions to validate your chipsets and the whole algorithm.

So we use generative AI to do a lot of things, including labelling and annotating the data and generating synthetic data. For this, we use some of the high performance chips by Nvidia because in the car you need to process a lot of these images, run the algorithm and get a decision within milliseconds. This is high performance compute on the edge. 

TCS used Nvidia’s Orin, Thor, those kind of systems-on-a-chip (SoCs) to do high performance compute on the edge. We use NVIDIA's generative AI models like NeMo and NIMS, which are microservices, to do this synthetic data generation.

The other big opportunity area is manufacturing. It’s a very critical opportunity for us to use generative AI.

With the advent of generative AI, especially on the shop floors, whether it is automotive manufacturing, chemical manufacturing, cement manufacturing, or chip manufacturing like electronics. All these areas you can actually start applying generative AI and create digital twins of the processes, assets and the working model itself. It could be at plant level, at a sub-process or a sub plant level or it could be at an equipment level.

In this area, Nvidia brings its Omniverse platform, which is a very good integration tool to simulate digitally the entire shop floor processes as well as the performance of the shop floor. So we are actually building competency on this particular platform.

TCS will be bringing the domain expertise in this partnership with Nvidia. We will bring the right software expertise and AI expertise to build those digital twins on top of Omniverse and work with customers in building those solutions.

For Gen AI, one area is around infrastructure (GPUs) and another is related to intelligence. Actually for intelligence, there will be a value increase. For infrastructure, the value will continuously drive down because cost of hardware will go down, cost of network will go down.

So compute will become cheaper but putting more intelligence, more value adding digital twins on an actual operation will be more valuable work for TCS because the impact is higher.

For example, if I build some kind of digital twin for a plant today, and then I check for the actual parameters I need to do to increase the efficiency by 3-4 percent. We did that with one of the power plants in Japan. So for the power plant which is 1000 megawatts, we could put one piece of AI-only digital twin for a very small portion for boiler optimisation. It improved their efficiency by 2 percent. But a 2 percent efficiency in such a large boiler decreases their fuel consumption, coal or gas consumption very significantly. It will save almost like $3 million-4 million per year for that one solution actually.

So imagine if you can take the entire power generation process and build maybe 15 digital twins and bring everything into a platform like Omniverse, that would have been very value-adding for that customer and they are willing to pay more for that.

We are going to continue to invest in some of this infrastructure. Depending upon how our Tata Group companies are also investing in Tata Communications and depending upon how things are shaping, we will definitely look at it. There is already partnership between Tata Communications and Nvidia.

In my group, for some of the latest chips that are coming in or are in the drive platform for Nvidia, we use those chips. We buy those reference chips regularly.

Today this (AI) is a technology. And the technology applies to multiple industries and it is very transformative. So if you look at it from my perspective of engineering, it can transform engineering itself, design engineering, construction. It can transform automotive industries. In the case of brownfield or greenfield plants, it can improve their operating efficiencies. And if it is a supply chain, it can really bring down the supply chain optimisation and the cost.

So there is a significant business impact for the use of this technology. As we see it getting adopted, we will see our business also proportionately increasing. It's hard to put a number to it now.