Please welcome Roland Busch to the stage.
I want you to think back to the time before electricity. The world moved at the pace of people, horses bridged distances, steam powered engines, and ideas moved only as fast as a letter or a human voice. Then electricity arrived, a general-purpose technology that turned night into day, amplified human capabilities, and powered progress. At Siemens, we helped build that world with electric currents, telecommunications, trains, and machines. One and a half centuries later, another general-purpose technology has arrived: artificial intelligence. AI will be as transformative for this century as electricity was for the last. It is revolutionizing how we design, build, and operate infrastructure, making systems, grids, cities, and economies more adaptive and efficient. Siemens helped build the world in the light of electricity, and now we do it again in the age of intelligence, powering the industrial AI revolution. It is moving faster than any transformation before: steam took 60 years, electricity 30, computers 15, and AI seven. When AI enters a physical system, it becomes a force with real-world impact. But AI that runs a production line, manages a power grid, or steers a logistics network must be reliable and safe. Hallucination is not acceptable in industrial AI. The solution is not to go more slowly but to embed AI technology from the beginning. In the next 40 minutes, we will show you the tools ready today to help you make progress with industrial AI. It starts with an idea. Ideas become designs, and from designs you start building. But long before you build, you want to know how it works. Siemens software helps you find those answers early with digital twins: physics-based, photorealistic, 3D, real-time replicas of whatever you want to create. You stress-test and break them virtually before spending a single euro. Our customers take it further with Digital Twin Composer, connecting digital twins of machines, production lines, and buildings into one integrated simulation. For example, a logistics company can test warehouse layouts virtually before moving a shelf. These simulations are hard, especially aerodynamics. With accelerated computing from NVIDIA, we drive calculations 10, 100, or even 1,000 times faster. Today at VivaTech, we announce new software in Simcenter 3D that eliminates meshing entirely, going straight from design to calculation in minutes. We also announce software collections for startups with up to 95% discount. For example, Latitude, a French startup building 3D-printed rocket engines, switched to our software and achieved a six-month speed-up and 15% higher efficiency. The next step is to engineer production. Manufacturing engineers create automation code on our TIA Portal, configuring thousands of components. This is repetitive and time-consuming. Less than two months ago, we launched the Eigen Engineering Agent, an industrial AI agent that completes engineering tasks end-to-end inside TIA Portal. It plans, collects data, writes executable code, compiles, and validates until it works. Customers report 50% higher productivity, 2.5 times faster development, and 80% higher quality. Today at VivaTech, we announce two new capabilities: it now understands electrical design plans and integrates them automatically, and it builds along common norms and frameworks. What used to take days now takes minutes. That's industrial AI: real impact, not hype. Now we bridge into the real world: production in a real factory. Industrial AI finds hidden potential in machines and processes. At an Audi plant, robot arms weld car bodies with millimeter precision. An industrial PC with a trained AI module on an NVIDIA GPU checks 2,000 welding spots per car body, inspecting, evaluating, and reporting. In Erlangen, Germany, we are turning a factory into the world's first fully AI-driven production site with 100 AI algorithms on the shop floor. Results: 40% faster time to market, 42% lower energy consumption, 69% productivity increase. We use physical AI to automate tasks that were out of reach at reasonable cost, like placing a plastic bag of screws. The tools are here. So why do many companies get stuck in pilot after pilot? Let's learn from companies that are succeeding. For this, I invite my fellow board member, Cedrik Neike.
Thank you so much, Roland. Very good. I will do a quick experiment to wake you up. Audience interaction: raise your hand if you use AI daily. Good. Now raise your hand if you are driving an AI project. Now raise your hand if your project is super successful and generating ROI. These few people, stand up. They are the heroes. Give them a round of applause. But the point is, we all talk about AI, we all use it, but we fail in industrial environments. I call it the three traps of adoption: first, the spray-and-pray syndrome—trying to shoot AI at everything, ending up with more pilots than Lufthansa. Second, the missing-foundation syndrome—having digital islands not connected. You need to build a foundation in your CRM, PLM, and ERP systems. At Siemens, we call it the data fabric. Third, the lonesome-cowboy syndrome—doing it on your own without partners. To discuss this, I want to bring Aiman Ezzat, CEO of Capgemini Group, to give an overview.
Yes. I mean, so basically, we run Capgemini, and we have a great partnership with Siemens for many years. Deploying industrial AI is more complex than using generative AI for productivity. It's about rewiring organizations and designing new business models with two key elements: new or existing products and services leveraging accelerated R&D, and new agentic business processes blending humans with AI agents. For that, we need strong foundations: technology and solid data fundamentals. Then deploying AI agents requires strong governance and control mechanisms, otherwise you have chaos and high token consumption. Finally, the human side: if humans don't trust the agent, it won't work. So that's the basis.
Absolutely. And you talk about trust: humans and AI agents trusting each other. We trust each other, and we're going to talk about two companies. Why are we actually working together? Can you give an overview of why we have such a good fit?
I think because it's a great complementarity. From your side, you bring technology, tools, and domain know-how in many industries. We bring strategy, transformation capabilities, end-to-end deployment in complex environments, and AI ecosystem management. We deploy together across industries from aerospace to consumer products to automotive to water. One client is GravitHy, building a green steel plant near Marseille. €2.2 billion investment, replacing coal with hydrogen. No pilot plant, so we built a digital twin of the factory and an operational AI layer to simulate and optimize hydrogen consumption in real time. It's a learning factory. Another example is Sanofi. In pharma, the challenge is scaling across many sites in a regulated industry. We developed Smart Operations, a manufacturing execution system that guides step by step, holds electronic batch records, and reduces deviations by 80%. The impact is concrete: AI brings something to the bottom line.
It makes a huge difference. Right? But let's stay in France and look at another company, like Sanofi. So Sanofi, you know, GravitHy is about reinventing an industry. And of course every industry has its own challenges. In pharma, the mission is to find the best and safest medicine, but getting there is complex. The challenge is scaling across many sites in a regulated environment. Sanofi has Smart Operations, a manufacturing execution system that is the cornerstone. It shows what needs to happen, records what happens, guides step by step, and holds electronic batch records. Instead of searching through documents, teams see the information they need. Siemens is working with Sanofi to roll that across the world. Why is that important? Moore's Law gives exponential cost advantage, but in pharma we have Eroom's Law: every nine months the cost of developing a new drug doubles. So we need AI and digital twins to link R&D and production. The impact: 70% reduction and 80% fewer deviations. So AI brings something to the bottom line. We talked about two industries, two companies. What did they do? They did not have a thousand AI projects. They built the foundation: digital twins and simulation, or the foundation to run factories and R&D. And they went with partners. The takeaway: it becomes easier if you focus, build the foundation, and partner. Aiman, thank you so much. Thank you for joining us today. I'm going to go back to Moore's Law and ask Roland to come back on stage with a very special partner.
Well. Two industries, two companies, one winning formula. They did not go for 100 pilot projects. They built the foundation and went with partners. That's the takeaway. Every AI success story you just heard runs on compute. Compute means chips, and chips need a home: energy-hungry data centers. Together with NVIDIA, we've developed the blueprint for the next generation of AI factories. We optimize energy use and cooling for the lowest cost per megawatt using digital twins of every subsystem. We use Digital Twin Composer to simulate the entire AI factory before a single cable is laid. Once built, you need to protect it. A single power fault can destroy millions worth of equipment. AI factories run on direct current (DC) end-to-end to avoid conversion losses. Our new circuit breaker interrupts faults in milliseconds, a thousand times faster than conventional systems. All of this is only possible with chips. Making those chips is a complex engineering challenge, and one company the entire world depends on is ASML. They build machines the size of houses that create the smallest structures humans have ever built. Please welcome on stage Christophe Fouquet, the CEO of ASML.
Hi, Christophe. Oh, you might have. You better. You're going to sit here. These are the biggest—maybe the most complex—machines I know. They're writing the smallest structures. We talk nanometers. You use light which doesn't really exist in natural light. Amazing technology. Do you ever think: holy shit, how does this whole thing run?
Well, we have to, because we have hundreds of those machines across the world. They have to print those AI chips every day and night, running 95% of the time. If you had to run your car 95% of the time, you'd be stressed. So it's a very important thing. Part of the complexity is making sure the machine can run without too many failures. Running them with highest precision and uptime is key. Some people say they're very expensive, but I say it's value for money. Moore's Law works only if the machines are working day and night.
That's right, that's right, that's right. So the economy has to work. One more reason to keep them running.
Because when the machine goes down, it's a major economic issue for our customers. How many components? Roughly more than 20,000. So this is another way to look at the complexity. You talked about the size, but the number of parts is very important because a lot of them were designed specifically for these machines. The technology is unique.
