Background: Epidemiological studies report evidence for an association between folate, an essential B vitamin, and the risk of several common cancers. However, both protective and harmful effects have been reported, and effects may differ by cancer site. These associations suggest that modulating dietary folate, or its synthetic form folic acid, could be used to modify population-wide cancer risk. However, observational studies are liable to biases, including residual confounding and reverse causation, thus limiting causal inference. Using Mendelian randomisation (MR), we investigated the causal relationships of genetically determined serum folate with pan-cancer risk (all cancers excluding non-melanoma skin cancers); breast, prostate, ovarian, lung, and colorectal cancers; and malignant melanoma. Methods: Using publicly available genome-wide association study (GWAS) summary data, we identified genetic instruments to proxy serum folate levels and analysed these using GWAS summary statistics of risk of pan-cancer and six site-specific cancers available from large consortia and the population-based cohort study UK Biobank (UKBB) within a two-sample Mendelian randomisation framework. We conducted MR using the inverse variance weighted (IVW) method and the likelihood-based approach. We performed sensitivity analyses to assess potential violations of MR assumptions. Results: We identified three SNPs (rs1801133, rs1999594, rs7545014) robustly associated with serum folate in a healthy, young adult Irish population using publicly available GWAS summary data. There was little evidence that genetically increased serum folate was associated with risk of pan-cancer or six site-specific cancers. Meta-analysis showed odds ratios (OR) per standard deviation (SD) increase in log10 serum folate of 0.92 (95% confidence interval 0.78-1.07) for breast cancer, 0.87 (95% confidence interval 0.71-1.06) for prostate cancer, 0.87 (95% confidence interval 0.61-1.25) for ovarian cancer, 0.87 (95% confidence interval 0.57-1.34) for lung cancer, and 1.26 (95% confidence interval 0.84-1.88) for colorectal cancer. ORs for pan-cancers and malignant melanoma in UKBB were 0.86 (95% confidence interval 0.71-1.03) and 0.57 (95% confidence interval 0.30-1.10) respectively. The results were powered to detect modest effect sizes (>80% power [α=0.05] to detect ORs 1.1 (or its inverse 0.9) for the cancer GWAS consortia) and were consistent between the two statistical approaches used (IVW and likelihood-based). Conclusions: There is little evidence that genetically increased serum folate may affect the risk of pan-cancer and six site-specific cancers. However, we may still be underpowered to detect clinically relevant but smaller magnitude effects. Our results provide some evidence that increasing levels of circulating folate through widespread supplementation or deregulation of fortification of foods with folic acid is unlikely to lead to moderate unintended population-wide increase in cancer risk.