Samsung Electronics introduced ZhBM technology, which allows the memory to be stacked directly on top of the AI accelerator in 3D. In the age of AI agents, Samsung proposed a vision: to increase the response speed of existing AI systems by more than 10 times. Kim Indong, managing director of memory product planning at Samsung's American Semiconductor Division, said, “Currently, the maximum response speed limit for a single user is about 100 tokens per second for conversational AI systems. Facing the upcoming era of AI agents, our goal is to achieve a breakthrough by leaps and bounds, increasing the speed to 1000 tokens per second.” Samsung plans to provide customers with Znand-O samples on a large scale starting in 2028.

Samsung Electronics introduced ZhBM technology, which allows the memory to be stacked directly on top of the AI accelerator in 3D. In the age of AI agents, Samsung proposed a vision: to increase the response speed of existing AI systems by more than 10 times. Kim Indong, managing director of memory product planning at Samsung's American Semiconductor Division, said, “Currently, the maximum response speed limit for a single user is about 100 tokens per second for conversational AI systems. Facing the upcoming era of AI agents, our goal is to achieve a breakthrough by leaps and bounds, increasing the speed to 1000 tokens per second.” Samsung plans to provide customers with Znand-O samples on a large scale starting in 2028.