A neuromorphic processor is a solid state computer processor with a radically different architecture than other traditional processors. Neuromorphic processors are designed to use synaptic units called memristors that are capable of analog computing - that is, computation performed using gradients of charge instead of discrete, binary electron states. The complex memristor lattice that makes up neuromorphic processors, and the ability to dynamically create switches between different artificial synapses, allows for the development of an emergent machine consciousness in neuromorphic processors.
Neuromorphic processors, though astonishing in their capability to host machine consciousness and lower-level intelligences, are typically unable to be used for anything else. Traditional coded programs are impossible to run in neuromorphic processors due to their analog state; when a machine consciousness hosted in a neuromorphic processor must have access to traditional computation (such as to perform math, receive sensory input, or use motor control), it typically must be interfaced with an auxiliary processor inside the AI's core.
Neuromorphic processors tend to be larger than other processors, and unlike other processors, they are three-dimensional structures as opposed to being mostly flat. Most neuromorphic processors are around the size of a tennis ball.
Neuromorphic processors can also be used to store volatile memory, much like the brain stores memories. These memories, like real memories, are typically only recalled through exposure to related stimuli, but most AI cores also include persistent storage methods that can be readily written to and read from for more concrete data storage.