For 178 years, Siemens has ridden—and often defined—epochal shifts in industry, from electrification to automation. Now, it is betting on a transition that may eclipse both. Dirk Didascalou, Head of Foundation Technologies, Siemens AG frames artificial intelligence not as incremental innovation but as a civilizational pivot: “There was a life on the planet Earth before electricity and after. There will be a life in industry before and after AI.” That framing is not rhetorical flourish. It underpins a deep internal re-architecture aimed at turning Siemens into a software-first, AI-native industrial company.
At the heart of this shift is “Foundation Technologies,” a 13,000-engineer organization that functions as Siemens’ technical spine. It consolidates four traditionally fragmented domains—research, platforms, product development, and user experience—into a single operating layer. The intent is both simple and radical: enforce a common foundation so every Siemens product is “open, easy, interoperable.”
As Didascalou explains, “We had to make this change to force this through ourselves,” acknowledging that the transformation is as much internal discipline as external promise. The structure mirrors modern tech companies, where platform teams serve internal developers as their primary “customers,” standardizing tools, services, and developer experience across the enterprise.
But the real disruption is economic and cultural. Industrial incumbents have long competed on physical assets—plants, machines, and long depreciation cycles. AI inverts that model. “The moment you go data-driven, everything changes,” Didascalou says. Value migrates from owning assets to extracting insight from data, compressing decision cycles, and continuously optimizing outcomes. The shift cascades across the enterprise: products become software-defined; revenue models move to subscription and usage; even finance functions struggle to adapt.
“Finance departments can’t even budget this… we say, it depends on what you use,” he notes, illustrating how AI-era business models destabilize traditional planning assumptions. What looks like a technology transition is, in reality, a full-stack reinvention of how industrial companies operate.
Crucially, Siemens is drawing a sharp line between consumer AI and industrial AI—an often-muddled distinction in boardroom conversations. Consumer systems predict human behavior; industrial systems act in the physical world, where errors carry real-world consequences.
“If the machine is hitting the wall, you can’t reset this. If it’s hitting a person, even worse,” Didascalou says. That raises the bar for safety, explainability, and determinism. It also changes the data equation. Machines don’t “speak” natural language; they generate time-series signals, CAD models, and process diagrams—modalities largely absent from the public internet.
The result is a fundamentally different AI stack: domain-specific models, proprietary data pipelines, and rigorous controls around IP and usage. Siemens’ approach is pragmatic rather than doctrinaire—combining large language models for intent with smaller, specialized models for execution. “A hammer is a good tool, but not for everything,” he says, capturing the company’s multi-model strategy.
This pragmatism extends to data—a subject often clouded by anxiety in industrial circles. The narrative of “dirty data” and “data scarcity,” Didascalou argues, misses the point. “All data is dirty… you can’t beat the mathematics,” he says, pointing to the iterative reality of model training, annotation, and refinement. More pointedly, he reframes the debate from protection to participation: “The question is not why should a customer share data. The question is what happens if you don’t.”
In an ecosystem where models improve with scale and diversity, opting out is not a neutral choice—it is a competitive disadvantage. The implication is stark: industrial AI will reward those who contribute to shared intelligence layers while safeguarding IP through architecture, not isolation.
Perhaps the most consequential break from the past is Siemens’ embrace of openness. For decades, industrial leaders thrived on closed systems that locked customers into vertically integrated stacks. AI renders that strategy obsolete. “You need access to all of the data… closed systems don’t work,” Didascalou says. Siemens’ answer is an “AI operating system” not as a monolithic standard, but as a set of interoperable capabilities—APIs, services, and emerging agent frameworks—that allow software and machines to collaborate across boundaries. It is a shift from owning the stack to orchestrating it. In this model, differentiation comes not from exclusion, but from how effectively a company enables ecosystems.
The broader signal is unmistakable. Industrial AI is not a feature upgrade; it is a re-foundation. Siemens is attempting to compress two decades of software evolution into a unified, AI-centric core—backed by sustained investment, acquisitions, and a willingness to rethink long-held assumptions about products, platforms, and partnerships. Whether it succeeds will depend less on any single technology and more on execution at scale. But the direction is clear. As industry moves from assets to algorithms, from closed systems to open ecosystems, the winners will not be those who adapt at the edges—but those who rebuild at the core.