We develop a machine-learning-based crop model emulator with the inputs and outputs of multiple global gridded crop model ensemble simulations to capture the year-to-year variation of crop yield under future climate change. The emulator can reproduce the year-to-year variation of simulated yield given by the crop models under CO₂, temperature, water, and nitrogen perturbations. Developing this emulator can provide a tool to project future climate change impact in a simple way.