Generative AI has transformed the way knowledge work is performed. In manufacturing industry, however, the most important questions are not about generating text or images but about understanding materials, manufacturing processes, and production systems. These challenges can be addressed through Physical AI, where artificial intelligence is combined with simulation and mathematical modelling to understand and predict the behaviour of materials, manufacturing processes, and industrial systems.
At Indalgo, we have been researching, developing, and delivering these kinds of solutions for the manufacturing industry for more than 25 years. In this article, we explain what Physical AI means in practice and why we believe it will become one of the most important competitive advantages in modern manufacturing.
For many people, AI today primarily means generative AI that helps search and analyse information, write content, create images, support software development, and assist with other knowledge-intensive tasks. Manufacturing companies, however, face a different set of questions: How do material properties evolve during a manufacturing process? How will a process change affect product quality, energy consumption, production costs, or delivery reliability? How can production processes be optimized? Can the impact of new materials, manufacturing methods, or production investments be evaluated before physical experiments or investment decisions are made?
A general-purpose language model can help explain these phenomena and make use of existing knowledge, but it does not understand the unique materials, manufacturing processes, or production environment of an individual company. Physical AI complements generative AI by combining artificial intelligence with company-specific data, physics-based models, simulation, optimization, and digital twins. As a result, decision-making can be based on the actual behaviour of a company’s materials, manufacturing processes, and production systems rather than on general assumptions.
When Indalgo was founded in 2010, its name was derived from the words Industry of Algorithms. The name reflects our belief that the competitiveness of manufacturing is built on the ability to understand, model, and continuously improve production systems, manufacturing processes, and materials through advanced algorithms.
Before the the company was established, Perttu Laurinen and Ilmari Juutilainen, whose backgrounds are in mathematics, statistics, and computer science, worked closely together at the University of Oulu. Their research focused on neural networks, data mining, and the enormous volumes of data generated by industrial processes, exploring how these emerging technologies could improve industrial operations. Their work advanced the development of software tailored for manufacturing environments and demonstrated how entirely new AI technologies could solve complex industrial challenges. This research ultimately laid the foundation for Indalgo.
The same curiosity, values, and fundamental questions continue to guide our work today. Our central question remains unchanged: How can the behaviour of the physical world be modelled accurately enough to enable better decisions? Today, this approach is widely described through concepts such as Digital Twins, Industrial AI, and Physical AI.
Manufacturing companies continuously develop new products, materials, and production processes. Traditionally, however, developing new materials, improving manufacturing processes, or making production investments has relied heavily on physical testing, with the true effects becoming visible only afterwards. The more complex the material, manufacturing process, or production system, the more difficult it becomes to predict how individual changes will influence the final outcome.
Physical AI allows a significant portion of this evaluation to take place digitally through simulation. When the behaviour of materials, manufacturing processes, and production systems can be modelled with sufficient accuracy, different alternatives can be simulated, compared, and optimized before physical experiments are conducted or investment decisions are made.
A good example is the microstructure of steel, which largely determines properties such as strength, toughness, and formability. This microstructure cannot be directly observed or measured during production. Instead, it is typically verified later through laboratory analysis of material samples under a microscope. Physical AI makes it possible to model this invisible world using production process data together with available measurements. This enables manufacturers to estimate the resulting microstructure and material properties long before laboratory results are available, or even before the material has been produced.
Indalgo has been developing these types of solutions together with manufacturing companies for more than 15 years. Our long-term collaboration with SSAB, for example, has included predicting material properties, modelling manufacturing processes, running simulations, and supporting industrial decision-making. Within the steel industry, this approach is often referred to as virtual steel production, where a digital representation of the steel manufacturing process supports research, product development, and continuous production improvement.
When the behaviour of the physical world can be predicted accurately, products themselves can also be designed differently. Traditionally, manufacturing starts with a raw material and progresses toward a finished product. Physical AI enables the opposite approach: starting from the desired end result and determining which material, manufacturing route, and production process will achieve it most effectively.
This approach is known as inverse design for manufacturing. Instead of asking what can be achieved with an existing material, the question becomes: What material and manufacturing process are required to achieve the desired properties? In this way, Physical AI not only optimizes existing production processes but also supports entirely new approaches to designing materials, products, and manufacturing solutions.
Developing Physical AI requires close collaboration between industry, universities, and research organizations. A good example is the €23 million Business Finland-funded RIS4E (Revolutionary and Intelligent Steel Solutions for Sustainable Environment) programme, where Indalgo is developing AI-, modelling-, and simulation-based solutions together with SSAB, John Deere Forestry, Konecranes, Rauma Marine Construction, Nomen, the University of Oulu, LUT University, and Tampere University.
For Indalgo, participating in the RIS4E programme is a natural continuation of a journey that began more than 25 years ago in academic research. Although today’s engineers have access to vastly greater computing power, more data, and significantly more advanced AI technologies than ever before, our objective has remained the same: to build a deeper understanding of the physical world and use that understanding to make better decisions.
Artificial intelligence is evolving rapidly, but we believe that the most important competitive advantage in manufacturing will not come from AI alone. It will come from combining AI with a deep understanding of each company’s physical world, including its materials, manufacturing processes, and production systems, and transforming that understanding into better decisions.