Panasonic Holdings Unveils KMC for Physical AI Reference Models

On November 11, 2026, Panasonic Holdings announced a Manufacturing Commons initiative, project name KMC, to develop AI and robotics reference models in Japan through industry, government and academic collaboration and deploy them across workplaces.

KMC is part of a broader effort to accelerate field-driven Physical AI, applying AI to physical workplaces by combining it with manufacturing expertise. The company wants to improve research, design and production processes as AI use extends beyond digital environments.

The proposed learning loop starts with AI learning from workers’ experience, skills and judgments. Workers use the resulting analysis and proposals, and the new knowledge and data generated through that work return to the AI for further learning. Panasonic’s goal is continuous, two-way learning between people, workplaces and AI across research, design and manufacturing settings.

Under KMC, organizations would share manufacturing technologies, knowledge and human connections across corporate and organizational boundaries. Panasonic’s stated ambition is to pool Japanese manufacturing capabilities into a globally competitive production model.

The announcement also points to the Technology Division’s 2040 technology vision, which pairs technology development with partner co-creation. It identifies AI and cyber-physical systems, which connect digital analysis with physical operations, as common technologies across research areas including robotics, sensing and model-based approaches.

Manufacturing Data and Process Applications

The strategy draws on more than a century of manufacturing experience in materials, devices, sensing, control and production processes. It also uses skilled workers’ techniques and methods for turning physical phenomena into usable data, giving the AI effort a base in factory operations as well as software.

One manufacturing area named in the announcement is glass-type perovskite solar cells. Panasonic’s materials technology and inkjet coating, laser processing and sealing techniques from display manufacturing enable power-generating glass for windows and walls. The company plans to use informatics and AI for materials development and process-condition optimization; business activity began with sales of pre-mass-production units in September 2026, and full mass production is targeted for April 2029 at the Nagahama mass-production technology development site in Shiga Prefecture.

The semiconductor work focuses on advanced back-end equipment for three-dimensional chip stacking and other next-generation manufacturing uses. It combines plasma surface modification, precision bonding, printing, inspection and transport technologies. A newly started AI Semiconductor Packaging Project unites research and business development, with AI analysis of process data and inspection images intended to support quality prediction, process-condition searches and equipment specifications.

T-Rex Provides a Robotics Example

A separate Panasonic research release dated November 5, 2026 shows the kind of robotics work behind the strategy. Panasonic Holdings and Panasonic R&D Company of America developed T-Rex with researchers from UC Berkeley, NVIDIA, Stanford University, Sapienza University of Rome and ItalAI. The system integrates fingertip tactile information into a vision-language-action model, which turns images and natural-language instructions into robot actions, allowing movements to be corrected as contact conditions change.

T-Rex generates its broad actions through a vision-and-language module running at about 5 Hz, or five updates per second, while a tactile module corrects motion at about 20 Hz. This allows responses to slipping or over-pushing without repeating slower image recognition for every adjustment. The T-Rex paper documents a 100-hour tactile-rich dataset and the variable-rate Mixture-of-Transformers architecture that accommodates these different processing rates.

Panasonic reports a 65% average success rate across 12 real-world contact tasks, compared with 35% for its strongest comparison method and 42% when tactile information was removed. The evaluation included page turning, transferring eggs, applying toothpaste, unlocking locks, removing cards and screwing in lightbulbs, requiring delicate force control and work with deformable objects.

Panasonic targets future uses including inserting, removing and tightening components in manufacturing and handling flexible or fragile items in logistics, with the goals of reducing worker burden and easing labor shortages. KMC’s central test is whether shared manufacturing knowledge can help turn research results such as T-Rex into reusable reference models across production and logistics workplaces.