The paper introduces Equivariant Graph Neural Operator (EGNO), a novel method for modeling 3D dynamics by explicitly capturing temporal correlations. Unlike existing methods that only predict the next state, EGNO models the entire trajectory as a function over time, using equivariant temporal convolutions parameterized in the Fourier space. This approach retains the intrinsic SE(3)-equivariance while efficiently modeling temporal dynamics. Comprehensive experiments on various benchmarks, including particle simulations, human motion capture, and molecular dynamics, demonstrate that EGNO outperforms existing methods, achieving significant improvements in accuracy and data efficiency. The code for EGNO is available at <https://github.com/MinkaiXu/egno>.The paper introduces Equivariant Graph Neural Operator (EGNO), a novel method for modeling 3D dynamics by explicitly capturing temporal correlations. Unlike existing methods that only predict the next state, EGNO models the entire trajectory as a function over time, using equivariant temporal convolutions parameterized in the Fourier space. This approach retains the intrinsic SE(3)-equivariance while efficiently modeling temporal dynamics. Comprehensive experiments on various benchmarks, including particle simulations, human motion capture, and molecular dynamics, demonstrate that EGNO outperforms existing methods, achieving significant improvements in accuracy and data efficiency. The code for EGNO is available at <https://github.com/MinkaiXu/egno>.