Hannover Messe 2025 shows the evolution of industrial AI – from GenAI via agentic AI to physical AI and industrial foundation models

The evolving AI vision – This year, AI was again the core topic at Hannover Messe, but discussions show that the topic keeps evolving. While GenAI was the dominant AI topic in previous years, agentic AI (autonomously acting software bots) was at the top of the agenda in 2025. We also got an idea of what the next big thing in the industry will be —physical AI (autonomously acting robotics systems). However, companies should not consider these topics in isolation, as workflows will integrate them tightly. Here is a quick example of what these workflows will look like: GenAI provides a convenient communication interface for users (via voice) to interact with an orchestrator agent. The orchestrator agent assigns tasks to many other AI agents. Depending on the request, they work in parallel or in sequence to perform many different and even complex, multi-step tasks. While AI agents with specific capabilities perform only digital tasks (e.g., booking a hotel), AI-based physical agents (robots) perform tasks in the physical world (e.g., cleaning a table). Although it sounds like science fiction, companies are already working on these technologies today. At Hannover Messe 2025, the German company Igus, for example, presented its first humanoid robot, called Iggy Rob; the Canadian company Santuary AI showed a humanoid robot demo at the Microsoft booth; and the Chinese company Booster Robotics presented its humanoid robot Booster T1. We will very likely see many more humanoid robots at Hannover Messe 2026. E&Y anticipates a huge market emerging and believes that by 2050, hundreds of millions of humanoid robots will “live” on this planet, more than Europeans today. Companies like Accenture and NVIDIA demonstrated their commitment to physical AI and announced partnerships with customers like Schaeffler to further explore the potential (Accenture and NVIDIA had already announced a similar partnership with KION at CES 2025). In 2024, BMW and Mercedes confirmed they were trialing humanoid robots in their factories in the US. At Hannover Messe 2025, we also noticed that vendors had started to introduce the first versions of industrial foundation models that understand and speak engineers’ language. Siemens is working on this topic together with Microsoft, but their offering is still in the very early stages. Alep Alpha is another example in this context; the company introduced its Pharia industrial suite, a sovereign foundation model for industrial engineering processes.

AI challenges persist – However, despite many high-flying ambitions and visions around AI, we should not ignore the current reality: Right now, there are not many success stories for the use of GenAI in the industrial context, and RoI is often difficult to calculate. Siemens won the Hermes Award 2025 for its Industrial Copilot, and the company claims that 200 customers are using it today. Nevertheless, we would like to point out that the Siemens Industrial Copilot portfolio includes, to our knowledge, 5 different copilots, which means that they have fewer than 40 customers per offering on average—that’s not that many. As RoI is difficult to measure, it is a challenge to identify a solid business and pricing model. In addition, while the availability of in-house training data to build specific AI models can be a clear differentiator, the legal aspects of data usage and sharing with others remain a challenge and lead to complex terms and conditions. In addition, before manufacturing companies can benefit from agentic AI at scale, they must do their homework regarding data management. They must overcome data silos and democratize access to internal data for real employees and digital employees (AI agents). This is still a massive task for many of them. Physical AI sounds very promising, but we should not forget that this technology is still years away from broad adoption.

My personal highlight: a concrete customer case – My personal award for the most interesting presentation at Hannover Messe 2025 goes to Andreas Evertz, the CEO of Flender Gruppe. He gave an excellent overview of how AI and IoT are transforming the engineering, service, and sales of gearboxes at Flender today. The company has implemented a digital engineering process to develop individual gearboxes based on customer requirements and supported by automated design options. It also does predictive maintenance and proactive sales calls around its gearboxes in operations triggered by IoT-based bots. His final statement was: “Competition is not the main problem; it’s the limited focus on innovation and the resistance to change”. I couldn’t agree more.

My recommendation – Manufacturing companies should focus on their data foundation today to get ready for agentic AI tomorrow.