MLP-Mixer has newly appeared as a new challenger against the realm of CNNsand transformer. Despite its simplicity compared to transformer, the concept ofchannel-mixing MLPs and token-mixing MLPs achieves noticeable performance invisual recognition tasks. Unlike images, point clouds are inherently sparse,unordered and irregular, which limits the direct use of MLP-Mixer for pointcloud understanding. In this paper, we propose PointMixer, a universal pointset operator that facilitates information sharing among unstructured 3D points.By simply replacing token-mixing MLPs with a softmax function, PointMixer can"mix" features within/between point sets. By doing so, PointMixer can bebroadly used in the network as inter-set mixing, intra-set mixing, and pyramidmixing. Extensive experiments show the competitive or superior performance ofPointMixer in semantic segmentation, classification, and point reconstructionagainst transformer-based methods.