Plasma YKL-40 is high in patients with cancer and in individuals who later develop cancer. Whether YKL-40 is only a marker or indeed a cause of cancer is presently unknown. We tested the hypothesis that observationally and genetically, high plasma YKL-40 is associated with high risk of cancer. For this purpose, we performed cohort and Mendelian randomization studies in 96,099 individuals from the Danish general population. Plasma levels of YKL-40 were measured in 21,643 and CHI3L1 rs4950928 was genotyped in 94,568 individuals. From 1943 through 2011, 2,291 individuals developed gastrointestinal cancer, 913 developed lung cancer, 2,863 women developed breast cancer, 1,557 men developed prostate cancer and 5,146 individuals developed other cancer. Follow-up was 100% complete. Multifactorially and CRP adjusted hazard ratio (HR) for gastrointestinal cancer was 1.82 (95%CI, 1.16–2.86) for 96–100% versus 0–33% YKL-40 percentile category. Corresponding HR were 1.71 (0.95–3.07) for lung cancer, but insignificant for breast cancer, prostate cancer and other cancers. CHI3L1 rs4950928 genotype was associated with plasmaYKL-40 levels, but not with risk of any cancer category. For gastrointestinal cancer, a doubling in YKL-40 was associated with a multifactorially and CRP adjusted observational HR of 1.14(1.05–1.23) for gastrointestinal cancer, but a corresponding genetic odds ratio of 1.06(0.94–1.18). For lung cancer, corresponding risk estimates were 1.11(1.00–1.22) observationally and 1.01(0.84–1.20) genetically. For other cancer categories, observational and genetic findings were insignificant. This study shows that high plasma YKL-40 levels were associated with high risk of gastrointestinal and likely of lung cancer, but genetic high levels were not.