Background: KRAS-mutant non-small cell lung cancers (NSCLCs) have unique clinicopathologic and genomic features, and novel therapies are in development to target many KRAS-mutant tumor types. Whether KRAS amplification also defines a unique, potentially targetable, molecular subset of NSCLC is currently unknown. Methods: Clinicopathologic and genomic data were abstracted from multiple independent cohorts of pts whose cancers underwent tumor genomic profiling at the Dana-Farber Cancer Institute (Cohort 1), AACR Project GENIE v.13, The Cancer Genome Atlas (TCGA), and 212 other studies (Cohort 2). Patient (pt) outcomes were examined according to KRAS amplification status (KRAS amplified vs KRAS non-amplified). Whole transcriptome sequencing and quantitative proteomics data from TCGA and the Cancer Cell Line Encyclopedia (CCLE) were used to correlate KRAS amplification with KRAS RNA levels and protein expression. Results: Among 15,341 pts with NSCLC, KRAS amplification was identified in 355 (2.3%) cases (median number of copies [range]: 11 [7-87]); the prevalence was similar in adenocarcinomas and squamous cell carcinomas. Compared to KRAS non-amplified cases, pts with KRAS-amplified NSCLC were more likely to be men (52.4% vs 45.6%, p = 0.01), have a history of smoking (82.1% vs 64.2%, p < 0.01), and higher median pack-years (35 vs 23.5, p < 0.01). However, median age was similar between the two groups (66 vs 66, p = 0.88). KRAS amplification was also associated with higher aneuploidy (median fraction of genome altered 30.4% vs 14.3%, p < 0.01), tumor mutational burden (p < 0.01 across different platforms), and increased median PD-L1 expression (20% vs 5%, p = 0.01). Of the 355 NSCLC samples with KRAS amplification, 152 (43%) had no concurrent oncogene driver mutations. In unbiased mutation enrichment analyses, KRAS-amplified tumors were enriched in concurrent mutations in KRAS, PALB2, POLE, and SLC34A2 (q < 0.1), while KRAS non-amplified tumors were enriched for oncogenic EGFR mutations (q < 0.1). Transcriptomic and proteomic profiling from the TCGA and CCLE cohorts demonstrated that KRAS amplification was associated with significantly increased KRAS mRNA (p < 0.01) and protein expression (p < 0.01), compared to KRAS non-amplified samples. In pts with available clinical outcomes data in the combined cohort (N = 9,335), the median overall survival (OS) from the date of diagnosis was significantly shorter in KRAS-amplified vs KRAS non-amplified cases (adjusted HR: 1.40, p < 0.01). KRAS amplification was confirmed to confer significantly worse survival outcomes in pts with both KRAS wild-type (N = 6,695, adjusted HR 1.37, p = 0.02) and KRAS-mutant (N = 2,640, adjusted HR 1.34, p = 0.02) NSCLC. These results were independently replicated in Cohort 1 and Cohort 2. Conclusions:KRAS amplification defines a novel molecular subset of NSCLC characterized by distinct clinicopathologic and genomic features and worse survival.