The MCD dataset is a comprehensive multi-campus dataset designed to address various challenges in robot perception. It features a wide range of sensing modalities, including CCS and NRE lidars, high-quality IMUs, cameras, and UWB sensors. The dataset includes semantic annotations for 29 classes across 59k sparse NRE lidar scans, providing a novel challenge for semantic segmentation research. MCD is the first dataset to provide point-wise annotations for NRE lidar systems and includes continuous-time ground truth based on optimization-based registration of lidar-inertial data on large survey-grade prior maps. The dataset includes 18 sequences, over 200k lidar scans, 1500k camera frames, and high-frequency IMU and UWB data. It covers three university campuses across Eurasia, providing a wide domain coverage and diverse environments. The dataset includes semantic annotations for NRE lidar scans and is the first extensive dataset with semantic annotations for NRE lidar. The continuous-time ground truth allows arbitrary time and density sampling and is essential for high FPS gaming AR/VR devices. The dataset also includes challenges in perception, such as motion distortion, extreme lighting, glass reflection, and solar interference. The dataset provides extensive benchmarks for state-of-the-art algorithms, including lidar-inertial SLAM, visual-inertial SLAM, range-aided localization, and semantic segmentation. The results show that existing methods struggle with the challenges posed by the dataset, highlighting the need for robust and precise solutions. The MCD dataset is a valuable resource for researchers in the field of robot perception and AI.The MCD dataset is a comprehensive multi-campus dataset designed to address various challenges in robot perception. It features a wide range of sensing modalities, including CCS and NRE lidars, high-quality IMUs, cameras, and UWB sensors. The dataset includes semantic annotations for 29 classes across 59k sparse NRE lidar scans, providing a novel challenge for semantic segmentation research. MCD is the first dataset to provide point-wise annotations for NRE lidar systems and includes continuous-time ground truth based on optimization-based registration of lidar-inertial data on large survey-grade prior maps. The dataset includes 18 sequences, over 200k lidar scans, 1500k camera frames, and high-frequency IMU and UWB data. It covers three university campuses across Eurasia, providing a wide domain coverage and diverse environments. The dataset includes semantic annotations for NRE lidar scans and is the first extensive dataset with semantic annotations for NRE lidar. The continuous-time ground truth allows arbitrary time and density sampling and is essential for high FPS gaming AR/VR devices. The dataset also includes challenges in perception, such as motion distortion, extreme lighting, glass reflection, and solar interference. The dataset provides extensive benchmarks for state-of-the-art algorithms, including lidar-inertial SLAM, visual-inertial SLAM, range-aided localization, and semantic segmentation. The results show that existing methods struggle with the challenges posed by the dataset, highlighting the need for robust and precise solutions. The MCD dataset is a valuable resource for researchers in the field of robot perception and AI.