The impact of the UCP2 −866G>A and UCP3 −55C>T variants on prospective risk of type 2 diabetes was examined over 15 years in 2,936 healthy middle-aged men (mean age 56 years). Conversion to diabetes (n = 169) was associated with higher BMI, blood pressure, cholesterol, triglycerides and C-reactive protein. The hazard ratio (HR) for diabetes of a BMI >30 kg/m2 was 3.96 (95% CI 2.87–5.47). Homozygosity for the UCP2A or UCP3T alleles accelerated the onset of diabetes, with significant differences in risk of diabetes at 10 years (HR [95% CI] UCP2AA vs. GA+GG 1.94 [1.18–3.19], P = 0.009; UCP3TT vs. CC+ CT 2.06 [1.06–3.99], P = 0.03) but less so at 15 years (UCP2AA 1.42 [0.92–2.19], P = 0.1; UCP3TT 1.57 [0.87–2.04], P = 0.13). Men who were homozygous for both UCP2AA and UCP3TT (1.5% of men) had a risk for diabetes at 10 years of 4.20 (1.70–10.37), P = 0.002. These genotype effects were additive with obesity, and men with a BMI >30 kg/m2 and this genotype combination had a 10-year risk of diabetes of 19.23 [5.63–63.69], P < 0.0001. Functional promoter variants UCP2 and UCP3 increase the prospective risk of diabetes. Although the mechanism of the UCP2 effect is likely to be caused by increased expression in the pancreas and subsequent reduced insulin secretion, the mechanism of the UCP3 effect is currently unknown. Both effects are exacerbated by obesity.