Abstract Recent epidemiologic studies have examined the association of fish consumption with upper gastrointestinal cancer risk, but the associations with n-3 and n-6 polyunsaturated fatty acid (PUFA) subtypes remain unclear. Using the National Institutes of Health–AARP Diet and Health Study (United States, 1995–2011), we prospectively investigated the associations of PUFA subtypes, ratios, and fish with the incidence of head and neck cancer (HNC; n = 2,453), esophageal adenocarcinoma (EA; n = 855), esophageal squamous cell carcinoma (n = 267), and gastric cancer (cardia: n = 603; noncardia: n = 631) among 468,952 participants (median follow-up, 15.5 years). A food frequency questionnaire assessed diet. Multivariable-adjusted hazard ratios were estimated using Cox proportional hazards regression. A Benjamini-Hochberg (BH) procedure was used for false-discovery control. Long-chain n-3 PUFAs were associated with a 20% decreased HNC and EA risk (for HNC, quintile5 vs. 1 hazard ratio = 0.81, 95% confidence interval: 0.71, 0.92, and BH-adjusted Ptrend = 0.001; and for EA, quintile5 vs. 1 hazard ratio = 0.79, 95% confidence interval: 0.64, 0.98, and BH-adjusted Ptrend = 0.1). Similar associations were observed for nonfried fish but only for high intake. Further, the ratio of long-chain n-3:n-6 was associated with a decreased HNC and EA risk. No consistent associations were observed for gastric cancer. Our results indicate that dietary long-chain n-3 PUFA and nonfried fish intake are associated with lower HNC and EA risk.