Building an AI operating system for Industry by Dr. Peter Koerte - Innovation Day India 2026

Thank you, Sunil. Good morning everyone. I want to congratulate you on yesterday's T20 win against England. I was rooting for you. Our chairman in India told me to wear blue to support the India team. This gets into the keynote: we're talking about AI, digital twins, simulation, and the industrial metaverse. Now we're at industrial AI. I asked a large language model who would win the World Cup; it said India and New Zealand, with a 50% probability, then 60-40 on Sunday. I'll discuss industrial AI and how it changes the physical world: mobility, trains, factories, life sciences, semiconductor plants, intelligent infrastructure. India has a massive growth opportunity at 7% GDP growth. India will gain and contribute the most. I'll cover: the AI revolution in industry, why India has a unique position to benefit, and how much operations can improve. AI without data is nothing. There are 20 billion smart devices, generating up to a terabyte of data per hour in a factory, but 80% is unused. At Siemens, we combine real and digital worlds. Examples: Audi uses AI for spot welding quality, reducing testing and saving time and money. In water infrastructure, we cut losses by half using sensors. In buildings, we developed Comfort AI, reducing energy costs by a third. Industrial AI must be safe, secure, reliable, and trustworthy. Unlike consumer AI, it can't hallucinate. India is a major force in applying AI across industries. 80% of Indian enterprises already use AI. There's an adoption gap, but India's educated workforce will use AI the most. If GDP share of manufacturing goes from 15% to 25%, and we make it AI-native, India can leapfrog. Example: our Bangalore team developed an application for an oil and gas customer in a desert, using satellite images and weather data to predict sandstorms with 70% accuracy seven days ahead, saving 20% in costs. The Siemens Xcelerator platform makes technology available for industrial use cases. It's an open digital platform with hardware and software. We have nine partners in the exhibition. Siemens connects real and digital worlds. Every third machine in the world is controlled by a Siemens controller. We are the number one industrial software company with $11 billion in digital business, the 16th largest software company. Industrial AI needs domain data, not internet data. We have 300 petabytes of data but need more. We work with ecosystems and data alliances. We partner with Nvidia for GPU acceleration. Examples: digital reality viewer and digital twin composer for faster simulation. We need to scale from pilot to production, industry by industry. Welcome to the stage Vishal of Nvidia and Sanjit of Adverb Technologies.

Wow. Okay. So 10 to 20% faster, more productive, 90% capturing failure rates earlier.

That's in virtual times. Now Sanjit, tell us about Adverb, what you've founded, created, and how those numbers reflect on using Siemens and Nvidia technology.

Great, thank you Sanjit. Fastest-growing robotics company. Now Vishal, you mentioned unlocking power for next-generation manufacturing. How do we do that in India? What are the challenges and how can we overcome them?

Fantastic, great examples. To conclude, Sanjit, you've built quadrupeds and mobile robots. Tell us where you see industries and which markets are important as you go from India to the world.

Thank you so much, Sanjit. Thank you so much, Vishal, for joining us on stage.

What we've seen is many proof of concepts that don't scale. The key is to get to scale. You have to do it first customer backwards, industry by industry. Understand the key use cases per industry—electronics, battery manufacturing, automotive—and build applications that can scale, otherwise it's just another proof of concept.

That's fantastic. Great summary. Very good. Thank you so much. Thank you.

Very good. Okay. Thank you once again and I hand it over to Indu, our master of ceremony. What are we doing next?

I joked with Sunil. We had so many numbers today. I'm not sure how many you'll remember. But one thing I kept in mind were the arrows pointing steeply upwards. Whether data center deployments, renewable energy, airports, semiconductor plants, all show a steep increase. You rarely find that anywhere in the world. It's a super exciting place.

Thank you for the question. On ethics, it's different in industrial AI compared to general AI. We don't have the same biases or discrimination issues because AI gives recommendations on production scenarios. What we do need is trustworthiness. For data security, the model is trained on large datasets from various sources, including synthetic data, to be generalizable. Once deployed, fine-tuning happens on specific data, and we need to ensure cybersecurity with zero-trust measures to prevent hacking. It's a very important question.

Domain knowhow is quintessential; otherwise it won't work. Example: we predict door failures on trains using traditional and generative AI. You need to understand which parts fail and when. Same in factories: you need to know which machine parts break down, material flow. You cannot separate it. You can fine-tune the AI to your specific manufacturing process because every industry and site is slightly different. There will always be a human in the loop for the foreseeable future, even with agentic systems.