CEATEC 2026 special report; When artificial intelligence comes out of the screen and gives life to robots

CEATEC 2026 annual technology exhibition in Japan has ended; An event where humanoid robots, robotic arms equipped with artificial intelligence and systems to perform real tasks in work environments presented a more tangible picture of the future of automation. From the robots that performed rhythmic movements to the system that picked the necessary vegetables from the shelf upon hearing the request for ingredients for a meal, the common message of the exhibition was clear; The competition of artificial intelligence is gradually going beyond the production of text and images and reaching the ability to understand the physical environment and perform actions in the real world.

The event was held from October 13 to 16 at the Makuhari Messe Exhibition Center in Chiba Prefecture, near Tokyo. The organizers had announced 821 companies and organizations from Japan and other countries for this year's course; 338 collections were present at the exhibition for the first time and 246 startups, universities and research institutes also participated in it. The number of registered visitors reached 112 thousand 373 people in four days.

Despite the attraction of humanoid robots to visitors, the technical significance of this year's CEATEC was not limited to the display of theatrical movements. An important part of the technologies presented addressed a more fundamental problem: how to connect artificial intelligence to a system that sees the environment, recognizes objects, makes decisions, applies the appropriate force and reacts to environmental changes?

Robots move from demos to real tasks

One of the newsworthy images of the exhibition were humanoid robots that displayed rhythmic movements, including the performance of the Macarena dance. Along with these performances, examples of robots doing tasks such as picking goods on shelves were also shown. These demonstrations are fun on the surface, but from an industrial perspective, they point to a more important issue: the coordinated movement of the robot's body is only one part of the problem, and the real use of robots requires understanding the environment, planning the movement and adapting to changing conditions.

In an environment such as a warehouse, store or factory, things are not always in a fixed place; Their size, shape, and orientation change, and the robot must be able to assess the situation without relying on a predetermined path. For this reason, this year's shows should be viewed in the context of the industry's move away from traditional automation and toward more environmentally friendly robots; Robots that require less detailed and separate programming for each situation.

Of course, the presence of a robot at the exhibition and the successful execution of a task alone does not prove that the technology is ready for widespread commercial deployment. Work speed, error rate, maintenance cost, safety with humans and performance in long shifts are still more important criteria to evaluate the industrial maturity of these products.

"Grocery of the future"; When the bot takes orders with natural language

One of the practical examples of the exhibition was the joint project of the Japanese company Yaskawa Electric and Amazon Web Services (AWS), which was introduced as "Grocery of the Future". In this demonstration, MOTOMAN NEXT's AI-equipped robotic arm was combined with a generative AI system to initiate the food selection process from a spoken request.

In the example reported by the Japanese media ITmedia, the user wanted the system to suggest the necessary materials to prepare a work. The AI assistant would answer about the ingredients, coordinate the required quantity with the user, and after confirmation, the robotic arm would start picking the vegetables. In this way, a typical request in natural language was converted into several operational steps: understanding the purpose, having a conversation to clarify the details, confirming the order, and physically executing it.

The important technical point was that the robot did not rely solely on fixed coordinates to pick up each vegetable. According to the report of ITmedia, the system rechecked the environment with the camera every time to recognize the location of Sabzi; Therefore, a slight change in layout could be managed without having to completely fix the position of objects. A 3D camera on the wrist and a tool to grab objects were also part of this process.

This example shows that combining linguistic models with robotics can change the automation user interface. Instead of the operator working with a specialized interface or predefined program for each operation, he can express his purpose and the system turns it into a set of actions.

However, this demonstration should not be equated with a fully automated store or a stand-alone robot that performs all retail tasks. What was presented was a demonstration of a limited but meaningful chain: dialogue, selection of raw materials, and execution of physical action.

Mitsubishi Electric; Focusing on physical AI in the factory

Mitsubishi Electric was one of the companies that highlighted its approach to Physical AI this year. Before the start of the exhibition, the company announced that it will display technologies for applying knowledge accumulated in industrial environments and combining it with data and artificial intelligence at its booth in Hall 8. The stated goal was to develop systems that can understand the environment and make decisions and act safely under changing real-world conditions.

One of the examples was related to the industrial assembly process; An area where slight differences between parts or small changes in their position can make automation difficult. Mitsubishi Electric demonstrated how artificial intelligence can be used to control robotic arms to tighten screws; An operation that requires adjusting the force and movement angle, assessing the condition of the part and making a decision according to the conditions.

Another example was the use of a humanoid robot to safely and appropriately place products of different shapes and sizes on the shelf. The company announced that the demonstration was done in collaboration with external partners and aimed to explore ways in which the robot could learn and apply knowledge and experience related to a specific work environment.

This approach is different from classical automation. In traditional systems, the environment usually has to be largely standardized and the movements predefined. In the approach based on physical artificial intelligence, the goal is to make the system better able to cope with the diversity of parts, changing conditions and uncertainty. If such capabilities can be reliably and cost-effectively brought into production, the range of tasks that can be automated will increase.

From human motion data to robot training

One of the technologies in focus at the CEATEC 2026 awards was the ContactGlove3 Pro hand gesture tracking glove developed by Murata Manufacturing and Melt Interface Technologies. This technology was introduced in the excellence awards section of the exhibition and uses electromagnetic field technology to record finger movements with high precision; A method that does not require an external camera to view the hand directly. According to the official description of the award, the glove is designed to record data used in robot training, remote control and skill transfer and can work continuously for more than eight hours.

The importance of such technologies is that robot training is not limited to recording joint coordinates or movement path. To recreate a human skill, the system must have meaningful data about how the task was performed and its context. For example, a technician's hand movements while assembling a part can become data to train a robot; Especially if the movement information is related to the work stage and its execution conditions.

This is one of the main challenges of developing general-purpose robots: providing high-quality, real-world data from what humans do naturally. Hand tracking technologies, touch sensors and remote control systems can reduce some of this distance; However, the final success rate depends on the quality of the data, the training method and the ability to transfer skills to different environments.

Why has the concept of "physical artificial intelligence" become important?

Artificial intelligence in recent years is mainly known for models that generate text, image, sound and code. But robots need a different kind of ability: understanding the physical situation, predicting the outcome of the movement, controlling the force, detecting the obstacle and executing the action without endangering humans or equipment.

For this reason, physical AI has become the nexus of several fields: AI models, machine vision, sensors, motion control, simulation, and robotics hardware. CEATEC demonstrations showed that companies are experimenting with combining these components for applications ranging from factories and shops to service environments. The examples of Mitsubishi Electric and the collaboration of AWS and Yaskawa show two different but complementary paths: the first focuses on versatility in industrial processes, and the second focuses on transforming linguistic requests into physical operations.

However, it should not be concluded from these shows that humanoid robots will soon replace human labor on a large scale. Commercial applications depend on indicators such as cost, speed, durability, safety, repairability and stable performance in real environments. An exhibition can reveal the direction of development, but alone is not enough evidence to prove the readiness of a large-scale market.

CEATEC is no longer just an exhibition of electronic products

CEATEC was held for the first time in 2000 and at the beginning it was more like a digital and consumer electronics exhibition. But in 2016 the organizers announced a shift in its approach towards broader digital innovation; The path that today includes industrial technologies, artificial intelligence, Internet of Things, business transformation and cooperation between companies and research institutions. The main theme of the 2026 course is "Transformation; It was where companies, industries and society evolve.

This year, 194 exhibitors from 31 countries and regions attended the exhibition. Also, conference programs and side events were held alongside the exhibition sections. This year, CEATEC was once again held at the same time as Japan Mobility Show Bizweek; A collaboration that aimed to link digital technology with the field of transportation and mobility.

This structural change is important because technologies such as robotics are no longer a standalone industry. Industrial robot depends on sensor, communication, data processing and artificial intelligence; A shopping robot needs language models, vision and tools for grasping objects; And developing more adaptive robots will also be difficult without data and simulation infrastructure. From this point of view, exhibitions like CEATEC try to bring together companies that are not able to build all the components of a solution alone.

Conclusion: The next competition is not just about getting smarter

CEATEC 2026 presented a picture of a new phase in robotics development: a shift from programmed movements and demonstrations to systems that can assess the environment, understand human request and perform a task in the real world. The food-picking robot at AWS and Yaskawa's project, the more adaptable industrial arms at Mitsubishi Electric's show, and the human-skill recording tools for training robots each show part of this path.

But the main test will be done outside the exhibition hall. Can these systems work with acceptable cost, high safety and reliable performance in factories, stores and service centers? Will their training be repeatable in different environments? And can businesses measure the actual productivity they achieve?

The answers to these questions will determine which technologies pass the demonstration stage and become operational tools. This year's CEATEC message can be summed up as follows: the next stage of AI competition is not just about building models with better answers; It is building systems that can function correctly and safely in the real world.