Samsung Electronics and SK hynix are making strides in commercializing “On-sensor AI” technology for image sensors, aiming to elevate their image sensor technologies centered around AI and challenge the market leader, Japan’s Sony, in dominating the next-generation market.
At the “SK Tech Summit 2023” held last month, SK hynix revealed its progress in developing On-sensor AI technology. This technology embeds an image sensor onto an AI chip, processing data directly at the sensor level, unlike traditional sensors that relay image information to the Central Processing Unit (CPU) for computation and inference. This advance is expected to be a key technology in enabling evolved Internet of Things (IoT) and smart home services, reducing power consumption and processing time.
SK hynix’s approach involves integrating an AI accelerator into the image sensor. The company is currently conducting proof-of-concept research focused on facial and object recognition features, using a Computing In Memory (CIM) accelerator, a next-generation technology capable of performing multiplication and addition operations required for AI model computations.
Additionally, SK hynix has presented its technologies for implementing On-sensor AI, including AI software and AI lightweighting, at major academic conferences like the International Conference on Computer Vision and the IEEE EDTM seminar on semiconductor manufacturing and next-generation devices.
Samsung Electronics is also rapidly incorporating AI into its image sensor business. This year, the company unveiled a 200-megapixel image sensor with an advanced zoom feature called Zoom Anyplace, which uses AI technology for automatic object tracking during close-ups. Samsung has set a long-term business goal to commercialize “Humanoid Sensors” capable of sensing and replicating human senses, with a road map to develop image sensors that can capture even the invisible by 2027.
In October, Park Yong-in, president of Samsung Electronics’ System LSI Business, emphasized at the Samsung System LSI Tech Day in Silicon Valley, the goal of pioneering the era of “Proactive AI,” advancing from generative AI through high-performance IP, short and long-range communication solutions, and System LSI Humanoids based on sensors mimicking human senses.
The push by both companies into On-sensor AI technology development is seen as a strategy to capture new AI-specific demands and increase their market share. The image sensor market, which temporarily contracted post-COVID-19 due to a downturn in the smartphone market, is now entering a new growth phase, expanding its applications from mobile to autonomous vehicles, extended reality devices, and robotics.
According to Counterpoint Research, Sony dominated the global image sensor market with a 54% share in the last year, while Samsung Electronics held second place with 29%, and SK hynix, struggling to close the gap, barely made it into the top five with 5%.