How Siemens, Kion, and OHB Are Building Industrial AI Through Data Partnerships

And, as you can see on the screen, my name is Peter. I look after technology and strategy at Siemens, among many other things.

Well, first of all, these are great partners, and we get along very well. Second, it is all about trust. We have known each other for a long time, and together we want to master the major challenges we all face. Whether you are KION, OHB, or Siemens, we all face the same challenges. Geopolitics is obviously a major issue right now. Everyone is dealing with uncertainty in the supply chain; nobody knows what will happen next. Then there is demographic change. We cannot find enough people, and there is a massive shortage of skilled labor. There is also the need to become more resilient and resource-efficient, as well as more sustainable and energy-efficient. We all face the same challenges, and technology is clearly how we must address them. We solve them not by working harder, but by working smarter—by accessing all the available data and knowledge. I think that is what unites us. We want to succeed in our respective industries, and we want to do it more intelligently.

First of all, we love partners with really big challenges, dreams, and aspirations, as Kristina and Rob both indicated: building something in space and revolutionizing the supply chain. What you need for that is intelligence; you need to be smart. And that comes down to data. That is why this year's session is about how we build data ecosystems. The major difference between industrial AI and all the discussions about generative AI and large language models is that consumer-oriented large language models are based on language, as the name suggests. Language can usually be found on the internet, so those models are trained on material published online. In industry, can you imagine OHB saying, 'Here is my latest design for a satellite constellation,' and publishing all the 3D designs, manufacturing data, and documentation on the internet? Certainly not. Rob would not do that with KION's solutions either. The major difference in industry is the data challenge: how do you collect data from production, engineering, and design, and then build the next generation of frontier models for use in the physical world? The only way is to build open data ecosystems. That is why it is so valuable to have Rob and Kristina here and to establish partnerships in which we agree to exchange data, with everyone contributing their domain expertise and the right datasets. We at Siemens can then help bring this together in a way that encapsulates the knowledge and domain know-how in models that can be scaled universally. That is the key, because none of us can do it alone. Siemens cannot. Yes, we produce trains, substations, and other products, but we do not have enough data. OHB does not have enough data, and KION does not have enough data. Together, however, we can move forward. We can also apply the AI expertise we have built, develop models around the data, ensure that no one's intellectual property is violated, and then begin to scale. That is the only way to do it.

There is a great quote: 'The future is already here; it is just unevenly distributed.' Kristina and Rob are bringing the future here and making that distribution uneven. They are creating an unfair advantage in their respective industries because they are moving early. What we are showing here at HANNOVER MESSE is the first example: the Eigen Engineering Agent. It helps industrial automation scale faster because automation systems have to be programmed, and today there are not enough programmers. We can significantly accelerate programming in terms of both quality and time. This is industrial AI already in action, and we can show that it is trustworthy, secure, and reliable. So far, we have optimized one step: programming a machine. But that is still piecemeal. There are many steps, whether you are designing a satellite, a forklift, or an AMR. What we want to connect is design, engineering, production, operations, and maintenance. They are all part of what we call the digital thread, so that you can optimize everything in one go and make it work. This is what Rob was already referring to as the industrial foundation model. The industrial foundation model thinks end to end. If I have to engineer and design a component, I can already consider its material properties, layout, and design, as well as how to manufacture it, so I can optimize it in one step. That is why access to data from engineering, manufacturing, and operations is so critical, and why this can only be achieved through partnerships. I wholeheartedly believe that we in Europe have a major global advantage: our domain knowledge. If we combine that domain knowledge with excellent partners such as KION and OHB and build the industrial foundation model together in a trusted way, we can scale it much faster than anyone else.