Gasgoo Munich-For the past two years, industry discussion has focused on what AI can do: Can it assist with design? Enable predictive maintenance? Auto-generate code?
A succession of proof-of-concept projects has fast-tracked AI’s entry into manufacturing firms. Yet, cases where it translates into scaled productivity remain scarce.
At the recent Siemens Realize LIVE user conference in Greater China, held in Shenzhen, a clear verdict emerged from the proceedings:
Industrial AI is moving from an era of "storytelling" to one of "crunching the numbers."
Image Credit: Siemens
Liang Naiming, Chairman and General Manager of Siemens Digital Industries Software China, pinpointed three stark realities facing the industry in his opening address: AI application scenarios are vast, yet struggle to create tangible industrial value; corporate data reserves are ample, but engineering knowledge remains fragmented; and while single-point pilot projects abound, scaling them up remains difficult.
These challenges map the chasm between "what AI can do" and "how to make it work."
The answer Siemens offered at this conference may well chart a path for the entire industry.
The "Siemens Solution" for Industrial AI
"Comprehensive digital twin, full lifecycle intelligence, and adaptive capabilities." That is how Tony Hemmelgarn, President and CEO of Siemens Digital Industries Software, defined the core strategy of Siemens' industrial software portfolio at the event.
These three strategic pillars help enterprises bridge the digital and physical worlds, accelerating innovation and generating tangible business value even as they navigate complexity and constant change.
Building on this, Siemens is embedding AI deeper across its entire product portfolio, focusing on three vectors: faster engine speeds, smarter execution, and delivering trusted outcomes.
These are not empty slogans; they are backed by hard data.
"Faster engine speed" points to exponential gains in efficiency.
In Designcenter, AI has accelerated drawing generation by 50%, while Teamcenter’s part reuse feature has boosted efficiency tenfold. "When others are increasing their speed by 5,000 times, your comfort zone is no longer secure," Tony Hemmelgarn remarked. Kinetic Vision, a small and medium-sized enterprise, saw its speed surge 4,000 times after adopting Siemens’ solutions, achieving 98% accuracy.
"Smarter execution" represents the shift of AI from an "assistant tool" to a "working partner."
Designcenter Copilot, honed over seven years, can predict a user's next move with 94% accuracy and was used 500,000 times last month alone. "This is Siemens' Copilot, not Microsoft's," stressed Joe Bohman, Executive Vice President of PLM Products at Siemens. Meanwhile, Teamcenter Copilot gives every engineer secure, instant access to the enterprise's full knowledge base, transforming a library of millions of documents into an interactive, intelligent resource.
"Delivering trusted outcomes" touches on the central proposition of industrial AI: trustworthiness.
Consumer-grade AI can afford to make mistakes; industrial AI cannot.
Image Credit: On-site Photo
Tony Hemmelgarn emphasized that "AI in engineering must always be grounded in physical models and engineering authenticity, verified through simulation and real-world product performance data to ensure capabilities are both powerful and reliable." The core product carrying this philosophy is Intelligence Center X, an industrial AI orchestration software unveiled by Siemens at the conference. It connects enterprise data with industrial know-how, helping users build intelligent applications, AI agents, and automated workflows. A key capability is Graph Studio, which creates knowledge graphs to link data from PLM, ERP, and CRM sources, allowing customers to run statistical learning on full factory datasets. GSK utilized this functionality to cut workload in core tasks by 90%.
This confirms the view of Jornt Moerland, Senior Vice President for Asia-Pacific at Intelligence Center X: "What enterprises truly need is not more AI tools, but industrial AI that creates business value."
Notably, Siemens has also introduced the concept of Industrial Ontologies.
Inside every customer, there are PLM, ERP, and CRM systems: PLM holds hardcore engineering data, ERP contains supply chain info, and CRM stores customer details and warranty records. Siemens uses Industrial Ontologies to bridge these data points.
Liang Naiming summarizes the issue as "abundant data, but fragmented knowledge." Many Chinese companies have stockpiled vast amounts of R&D, manufacturing, and operational data, yet it remains siloed across different systems and business processes. The underlying engineering knowledge, industry experience, and business logic have yet to be fully connected and utilized.
This is precisely where industrial AI diverges from its consumer-grade counterpart.
For a general AI, whether an answer is good or bad is a matter of efficiency. But when AI participates in product design, engineering simulation, or manufacturing, it faces a highly complex, constrained reality that bears engineering responsibility. The first threshold for industrial AI has never been how smart the model is, but whether the enterprise has integrated AI into workflows that actually deliver value.
Siemens’ answer is clear: AI is not an isolated functional module, but a capability embedded throughout the digital twin, the digital thread, and the entire industrial software chain.
From the three strategic pillars to the three AI vectors, and finally to the vehicle of Intelligence Center X, Siemens has constructed a comprehensive framework for industrial AI implementation—stretching from strategy to product.
From BYD to Dongfeng: AI Implementation Case Studies in the Auto Industry
The automotive sector serves as the most convincing testing ground for Siemens' industrial AI strategy.
In the Chinese market, the logic of "expanding from a point to an area" is particularly pronounced.
At the conference, Tony Hemmelgarn cited a series of Chinese automotive clients: BYD is applying the full Siemens portfolio; Dongfeng Motor has deployed Teamcenter—and clients who start with Teamcenter typically expand to the full portfolio.
Image Credit: On-site Photo
The BYD case is particularly telling. During the conference, Tony Hemmelgarn revealed that he had recently visited BYD and witnessed them "widely using our product portfolio." From NX computer-aided design software on the design side, to the product lifecycle management platform Teamcenter, and on to simulation testing software Simcenter, BYD covers nearly the entire spectrum of Siemens' industrial software lineup.
Dongfeng Motor represents a different, yet equally typical, path.
Based on Tony Hemmelgarn's experience, "customers often start with Teamcenter and eventually expand to the full portfolio"—and Dongfeng fits this pattern precisely. As the core of full lifecycle intelligence, Teamcenter is one of the world's most mature and scalable PLM tools. It acts not only as the backbone of data management but as the foundation of the digital thread—linking data across design, manufacturing, and operations. After implementing Teamcenter, Dongfeng Liuzhou Motor improved data accuracy and cut model query time by 50%.
CATL offers a success story from yet another angle. The global battery giant has begun adopting Siemens' Polarion tool. Polarion is a requirements management platform for complex systems and software development; its adoption signals that Siemens' capabilities are extending from traditional mechanical design toward software-defined products—a core proposition for intelligent vehicles.
Beyond these leading players, Siemens' penetration in the auto sector is deepening. The SICAR white paper on standardization in the automotive industry, released in March 2026, shows that the SICAR automation standardization solution has been widely adopted by carmakers including BYD, XPENG, NIO, Xiaomi, BAIC, GAC, and Li Auto, as well as numerous line builders and component suppliers.
Why has the automotive industry become the densest deployment zone for Siemens' industrial AI?
The answer likely lies in the sheer complexity of automotive engineering itself.
Modern vehicles are no longer purely mechanical products. Software teams must begin coding before hardware is ready, creating an "integration nightmare" with traditional methods. Using tools like Capital, Expedition, and Simcenter, engineers can simulate and verify entire systems before chips are even taped out.
In his speech, Tony Hemmelgarn underscored the strategic importance of combining PLM with EDA (Electronic Design Automation): "The integration of PLM and EDA is not optional; it is dictated by the nature of the complex engineering problems we solve. Modern products integrate design, electronics, PCBs, and even integrated circuits—all elements must collaborate seamlessly."
Joe Bohman illustrated AI's value in the auto sector through a more specific scenario. Connector giant Molex once posed a question to him: "We plan to hire 1,000 engineers. Can AI help these new engineers quickly master your tools? Can it help them learn our way of doing things?"
This is the universal challenge facing nearly every company in the automotive supply chain: a massive talent gap and difficult knowledge transfer. AI has become the only viable path to break through.
From BYD's shortened R&D cycles to Dongfeng's Teamcenter deployment, and CATL's adoption of Polarion, these cases point to a single conclusion: in the automotive industry, Siemens' industrial AI is not a decorative "add-on" module, but a full-chain productivity tool spanning design, manufacturing, and operations.
From Humanoid Robots to AI Factories: The Industrial Path of Embodied Intelligence
If the automotive sector is the "main battlefield" for Siemens' industrial AI, then embodied intelligence is the "new frontier" currently being opened up.
Apptronik was a case highlighted by Tony Hemmelgarn. The humanoid robotics company replaced other vendors' products with Siemens' full software stack. Why?
"To simulate the entire system and achieve cost reduction, energy savings, and lightweighting, all components must be brought into a unified framework for collaborative design," Tony Hemmelgarn explained.
Even more critical, he noted, "we don't just simulate the robot itself; more importantly, we can simulate the entire process." Siemens has proven this capability over years of traditional robot applications in automotive welding and painting. Now, the company is migrating that same methodology to the realm of humanoid robots.
Image Credit: On-site Photo
This reveals a crucial industrial logic: humanoid robots are not essentially a new industry, but an extension of industrial robotic capabilities into the dimension of complex systems.
The process planning, path simulation, and system coordination capabilities Siemens has accumulated on automotive production line robots can be directly transferred to humanoid robot scenarios.
The strategic partnership with NVIDIA represents another key move for Siemens in the field of embodied intelligence.
At CES in January 2026, Siemens and NVIDIA announced a major expansion of their strategic partnership to jointly build an industrial AI operating system. The companies plan to create the world's first fully AI-driven, adaptive manufacturing bases globally, kicking off implementation in 2026 with the Siemens electronics factory in Erlangen, Germany, as the pilot.
A landmark result of this collaboration materialized in April 2026: Siemens, partnering with humanoid robotics company Humanoid and NVIDIA, successfully completed operational tests of humanoid robots at the Erlangen factory. The wheeled humanoid robot HMND 01 Alpha, equipped with NVIDIA's physical AI tech stack, autonomously executed logistics tasks such as tote handling.
In the Chinese market, Siemens is also accelerating its ecosystem layout for embodied intelligence.
During the 2026 World Artificial Intelligence Conference, Guanglun Intelligence officially partnered with Siemens. The two will connect Siemens' industrial simulation with Guanglun's SimFoundry platform, converting industrial engineering knowledge into a simulation world that robots can interact with, learn from, and be evaluated against—solving the industry-wide challenge of translating embodied intelligence between virtual and physical industrial environments.
Additionally, Siemens signed a strategic cooperation agreement with Songying Technology to collaborate on physical simulation, industrial digital twins, and embodied intelligence, jointly advancing the application of physical AI in advanced manufacturing scenarios.
From Apptronik's software stack switch, to the industrial AI operating system with NVIDIA, and the local ecosystem partnerships with Guanglun Intelligence and Songying Technology, Siemens' layout in embodied intelligence reveals a clear hierarchy:
Using industrial simulation and digital twins as the foundation, physical AI as the engine, and an open ecosystem as its reach, the company is pushing embodied intelligence from the laboratory into real-world industrial scenarios.
Tony Hemmelgarn’s summary at the conference perhaps best encapsulates Siemens' overall view on industrial AI: "The digital twin is opening up new business growth for us; lifecycle intelligence is accelerating Teamcenter's growth because AI has become a customer imperative; and the adaptive strategy ensures customers never get stuck in data silos, but grow with us."
As industrial AI moves from proof-of-concept to scaled deployment, and as the question shifts from "can AI do it?" to "how much value can it create?", the path Siemens has charted is essentially a closed loop: from data to knowledge, from knowledge to intelligence, and from intelligence to productivity. The journey begins with a comprehensive digital twin and ends with every industrial decision being made faster, more accurately, and with greater trust.
