A way to share AI work between different chips, like CPUs, GPUs, and NPUs.
It is like a school group project. The CPU plans. The GPU does the heavy lifting. The NPU saves battery. Nobody makes the class clown do the math.
It sends each job to the best chip. You meet it in phones, computers, and data centers.
GPU
Heterogeneous computing often gives big parallel AI jobs to a GPU.
NPU
Heterogeneous computing gives power-saving AI jobs to an NPU.
AI chip
An AI chip uses different processors to balance speed and power.