Sony Semiconductor Solutions and Taiwan Semiconductor Manufacturing Company have signed a non-binding memorandum of understanding to explore a strategic partnership for the development and manufacturing of next-generation CMOS image sensors. Under the proposed structure, the two companies intend to establish a joint venture in Japan, with Sony expected to be the majority and controlling shareholder.
The planned joint venture would set up development and production lines in Sony’s newly constructed fabrication facility in Koshi City, Kumamoto Prefecture, Japan. The goal is to combine Sony’s image sensor design expertise with TSMC’s process technology and manufacturing capabilities.
This is an important development for the semiconductor industry because CMOS image sensors are no longer limited to digital cameras or smartphones. They are increasingly used in automotive systems, industrial vision, robotics, medical devices and AI-enabled sensing applications.
A CMOS image sensor, or CIS, is a semiconductor device that converts light into electrical and digital signals. These sensors are used in applications such as smartphone cameras, automotive cameras, security systems, machine vision, medical imaging and industrial inspection.
The market is strategically important because image sensors sit at the intersection of optics, semiconductor manufacturing, packaging, software and artificial intelligence. As more systems require visual input, the demand for advanced image sensors is expected to grow.
Sony is one of the leading suppliers of CMOS image sensors globally. Its sensors are used in many high-end smartphones and imaging systems. TSMC, meanwhile, is the world’s largest pure-play foundry and brings advanced process technology, manufacturing scale and semiconductor production know-how.
The proposed Sony–TSMC partnership therefore brings together two very different strengths: Sony’s sensor architecture and imaging expertise, and TSMC’s foundry manufacturing capability.
According to Sony Semiconductor Solutions, the proposed partnership is designed to develop and manufacture next-generation image sensors. Sony and TSMC are also discussing potential investments by the joint venture, as well as additional Sony capital investment in its existing Nagasaki plant. These investments are expected to be implemented in phases, depending on market demand and government support from Japan.
The companies also said the partnership will explore opportunities in physical AI applications, including automotive and robotics.
Physical AI refers to systems that use sensors, processors and software to understand and interact with the physical world. Examples include autonomous vehicles, robots, smart factories, drones and machine vision systems. These applications require increasingly advanced sensors with better sensitivity, faster readout, lower noise, higher dynamic range and improved integration with AI processing.
The proposed Sony–TSMC image sensor joint venture also fits into TSMC’s broader expansion in Japan.
TSMC’s first Japanese fab, operated by Japan Advanced Semiconductor Manufacturing, or JASM, began volume production in late 2024. JASM was established with support from Sony Semiconductor Solutions, and Denso and Toyota later joined as minority investors.
The first Kumamoto fab focuses mainly on 12nm to 28nm logic chips for automotive and industrial applications.
TSMC is also moving forward with a second Japanese fab. According to a Taiwanese government filing reported by Reuters, TSMC is expected to launch equipment installation and mass production of 3nm wafers in 2028 at its second factory in Japan. The planned monthly capacity is around 15,000 12-inch wafers.
This makes Japan an increasingly important manufacturing location for TSMC, especially as global customers look for more geographically diversified semiconductor supply chains.
For Sony, the proposed joint venture could help strengthen its position in image sensors at a time when the market is becoming more demanding.
Smartphone camera systems continue to require higher performance, smaller pixels, faster readout and better low-light capability. At the same time, automotive and robotics applications require sensors with high reliability, high dynamic range and advanced sensing features.
By working more closely with TSMC, Sony may be able to accelerate the development of more advanced sensor technologies and improve its manufacturing flexibility.
This is especially important as image sensors become more complex. Modern CIS devices may involve stacked architectures, advanced logic integration, backside illumination, high-speed interfaces and specialized processing features. These requirements make manufacturing technology a key competitive factor.
For TSMC, the partnership deepens its relationship with Sony and expands its role in the image sensor value chain.
TSMC is already the dominant foundry for many types of logic, high-performance computing and mobile chips. Image sensors are a different category, requiring close integration between pixel design, analog circuitry, process optimization and manufacturing control.
A closer Sony–TSMC collaboration could help TSMC participate in a broader range of specialty semiconductor markets, including imaging, sensing, automotive vision and robotics.
It also strengthens TSMC’s position in Japan, where the company is already building a larger manufacturing footprint through JASM.
One of the most interesting parts of the announcement is the reference to physical AI, automotive and robotics.
Cars and robots increasingly rely on cameras and sensors to detect objects, understand surroundings and make decisions. These systems need sensors that can perform reliably in challenging conditions, including low light, high contrast, fast motion and harsh environments.
For automotive applications, image sensors must also meet strict reliability and quality standards. In robotics, sensors need to support real-time perception, object recognition, navigation and human-machine interaction.
The Sony–TSMC partnership could therefore support a new generation of sensors designed not only for image capture, but also for machine perception.
