Siemens CEO's mission to automate everything

What It Covers

Roland Busch, CEO of Siemens, explains how the 175-year-old industrial technology company operates across manufacturing, buildings, mobility, and energy systems. He details Siemens' organizational transformation to scale AI and automation horizontally across 320,000 employees globally, while navigating rising trade barriers and the shift from automating physical processes to automating knowledge work.

Key Questions Answered

• •Organizational Restructuring for AI Scale: Siemens implements a "one tech company" program creating horizontal fabrics across previously siloed divisions—data fabric, technology fabric, and sales fabric. This allows unified customer identification across business units, shared digital platforms, and consolidated AI capabilities. The transformation removes organizational layers, consolidates operations into six major units, and enables data aggregation essential for training industrial AI models without losing domain expertise in specific verticals.
• •Local Manufacturing Footprint Mitigates Tariff Impact: Siemens maintains 85-87% local content in major markets like the US and China, with 45,000 US employees and similar regional distribution globally. This localization strategy limits direct tariff impact to low-mid single digits on bottom line. The company doubles manufacturing capacity for products like digital switching in the US and invests in assembly lines domestically, though warns customers suffer more from trade barriers than Siemens itself.
• •Industrial AI Requires Domain-Specific Training: Generic large language models achieve only 60-70% accuracy for industrial applications, insufficient for manufacturing deployment. Siemens trains LLMs on proprietary design data, machine operation data, historical maintenance records, and synthetic data to reach 95%+ accuracy rates. Example: photorealistic ray tracing of digital parts versus standard digital representations dramatically improved robotic grip-in-box hit rates, demonstrating that granular domain data makes industrial AI viable.
• •Digital Twin Composer Closes the Automation Loop: Siemens builds comprehensive physics-based digital twins that ingest real-time data from manufacturing lines, environmental sensors, and machine drawings. The system simulates forward and backward in time, identifies problems, and deploys AI agents that act as trained supervisors—detecting issues, recommending fixes, and automatically updating software. Human workers receive guidance through AR glasses for physical interventions, creating a closed-loop system between digital simulation and physical operation.
• •Automation Economics in Aging Societies: Fully automated factories produce higher output with fewer workers per unit, which Busch frames as necessary for aging populations in Germany, Japan, Korea, and China facing steep demographic curves. He argues labor should shift to irreplaceable roles in healthcare and social services rather than manufacturing. However, this creates tension around job displacement, as automated facilities require substantial land and energy while generating limited employment compared to traditional manufacturing.
• •Data Sharing Alliances Enable Model Training: Nine top German machine builders including TRUMPF, DMG, and GILDEMEISTER share operational data with Siemens to train AI models, recognizing individual company data volumes prove insufficient. These manufacturers trust Siemens as a decades-long partner to aggregate data without sharing their latest machine designs. This collaborative approach creates data pools large enough to train models that enable autonomous machine operation where users simply specify the part and desired outcome.