We assembled a compendium dataset of mass-spectrometry-based proteomics data of 2002 primary tumors from 14 cancer types and 17 studies. Most of these previous studies were led by either CPTAC or the International Cancer Proteogenome Consortium (ICPC). The cancer types represented in the proteomics compendium dataset were the following: Breast Invasive Carcinoma (n=230 tumors with proteomics data), Colorectal Adenocarcinoma (n=187), Gastric Cancer (n=80), Glioblastoma (n=100), Head and Neck Squamous Cell Carcinoma (n=108), Hepatocellular Carcinoma (n=165), Lung Adenocarcinoma (n=111), Lung Squamous Cell Carcinoma (n=110), Ovarian Serous Cystadenocarcinoma (n=269), Pancreatic Ductal Adenocarcinoma (n=137), Pediatric Brain Tumors (n=219), Prostate Adenocarcinoma (n=76), Renal Cell Carcinoma (n=110), and Uterine Corpus Endometrial Carcinoma (n=100). The 2002 tumors in the compendium represented 1982 patients, with the pediatric brain tumor dataset involving 219 tumors from 199 patients. The above studies analyzed the tumors using liquid chromatography-tandem mass spectrometry (LC-MS/MS) global proteomic and phosphoproteomic profiling. Each molecular dataset is uploaded on GitHub as a series of separate project files. Each file uses a common protein feature set with the same ordering across files. One can concatenate the files together using Excel. Each molecular dataset has a common sample set, allowing one to derive correlations between datasets (e.g., between mRNA and protein). The phospho-protein datasets consist of 5419 phospho-protein features that had available data for >50% of samples in at least seven cancer types.