Siemens, with over 175 years of industrial experience, has officially partnered with NVIDIA to deepen their collaboration in the field of industrial AI. The two companies are not only integrating hardware and software, but also comprehensively combining AI, analog, digital twins, and automation, aiming to create an industrial AI operating system that can be "practically operational and scalably deployed" in real factories. Siemens CEO Roland Busch and NVIDIA CEO Jensen Huang provided their first comprehensive explanation of the implementation timeline, application scenarios, and substantial impact on manufacturing, energy, and the global supply chain of this industrial AI system.
Industrial AI is being implemented in factories, moving from assisting decision-making to handling tasks on behalf of employees.
Busch points out that the key change in industrial AI now is that the new generation of models not only provides suggestions, but can directly make decisions and execute on behalf of humans, enabling the system to begin to have autonomy and self-adjustment capabilities.
He also mentioned that many customers have gone further and created digital twins of their manufacturing processes, first optimizing the processes in the virtual world and then implementing them in the real factory; AI is indeed already operating on the production line, but it is moving towards a higher level.
Scaling up is the real challenge; the barrier to entry needs to be lowered to make deployment and replication easier.
Busch frankly admitted that the real difficulty lies not in the feasibility of AI, but in its ability to be scaled up and widely adopted. Key hurdles include:
- Do customers have sufficient skills?
- Is the system easy to deploy?
- And whether it can be quickly replicated across factories and industries.
Currently, the adoption of industrial AI still heavily relies on specialized personnel and complex integration. Therefore, Siemens is focusing on lowering the barrier to entry, making deployment easier and more intuitive. He also emphasized that related solutions have begun to be adopted in industries ranging from shipbuilding and heavy industry to startups, indicating that market momentum is rapidly building.
Nvidia accelerates Siemens software integration, streamlining the entire process from design to factory.
Jensen Huang stated that this collaboration is not a symbolic alliance, but a deep integration across software and hardware, and across processes. Key aspects of the collaboration include:
- Accelerate Siemens' EDA software
- Accelerated Physics and Process Simulation Software
- Integrating AI, physical AI, and large-scale models into Teamcenter and factory automation systems
This means that Nvidia will rely more directly on Siemens' analog and digital twin tools when designing chips and systems in the future; at the same time, Nvidia's own factories and partners (such as Foxconn) can also use this industrial AI operating system for production line and factory management, forming a complete closed loop from R&D to manufacturing.
(Note: Teamcenter is a product lifecycle management software developed by Siemens. It is a digital platform that connects people, processes, and data within an enterprise. Through a unified digital thread, it integrates mechanical, electronic, and software design, bill of materials, and process management, assisting enterprises in collaborating throughout the entire process from product concept and design to manufacturing and service, accelerating time-to-market and reducing development costs. EDA software uses computer-aided design (CAD) tools to automate the complex design process of integrated circuits (ICs) and electronic systems, covering logic design, circuit simulation, placement, verification, etc.)
Digital twins reduce trial-and-error costs, while edge inference accelerates efficiency.
When discussing the impact of AI on the real world, Jensen Huang cited the "Vera Rubin" as an example, explaining that the system's complexity and cost pressures have reached a point where a completely new design approach is needed. This system integrates six chips, with a single GPU consuming up to 240,000 watts of power, achieving 10 times the energy efficiency and cost efficiency of its predecessor.
His key point is that if the entire system design and verification can be completed in Siemens' digital twin, the cost of trial and error can be greatly reduced, turning the "impossible" into the "mass-producible," and getting closer to a one-step solution.
Busch added that the battleground for AI is not just in data centers; its real value lies in whether low-latency inference can reach the edge of the factory. AI chips are now being used in controllers, industrial computers, and edge devices, allowing factories to adjust and optimize in real time, rather than analyzing after the fact, thereby further improving yield, energy consumption, and overall efficiency.
(Note: Edge devices are computers/controllers installed in factories, on machines, or on-site that can sense, process, and react in real time.)
Autonomous factories and energy bottlenecks coexist, and supply chain pressures extend all the way to space.
Both sides agreed that demand for autonomous and highly automated factories is rising, driven by factors including labor shortages, improved yields, better energy efficiency, and is particularly crucial for the return of U.S. manufacturing.
Huang Renxun described modern factories as "giant robots," and the biggest challenge in the past was that robots were too difficult to teach and required too much human effort in software development. The value of physical AI lies in making robots easier to "teach," replacing a large amount of handwritten programming with demonstrations.
Regarding energy, Huang Renxun stated bluntly that all industrial revolutions are constrained by energy, and the AI revolution is no exception. Therefore, each generation of products must be more energy-efficient. Busch then broadened the perspective to the entire power supply chain, pointing out that the demand for high-quality electricity from data centers has put pressure on power generation, gas turbines, high-voltage transformers, and power distribution equipment, and bottlenecks may occur in some areas.
Extending the discussion to the Chinese market, Huang Renxun stated that demand remains strong, with attitudes largely reflected indirectly through enterprise channels. Busch also mentioned that Siemens' software investments will continue to expand, and mergers and acquisitions are not ruled out.
Finally, the two extended their vision to the longer term, suggesting that space data centers might possess advantages in energy and cooling. If production were to take place in space, the most suitable products would be intelligent and computational power capable of being rapidly transmitted back to Earth. In the next two to three years, with the full integration of AI, digital twins, and automation, autonomous factories will no longer be just a concept, but a new starting point for global manufacturing competition.
This article, "Siemens and NVIDIA jointly launch industrial AI system: from digital twins to autonomous factories, accelerating the implementation of AI in manufacturing," first appeared on ABMedia, a ABMedia .
