We studied the effect of birth size on glucose and insulin metabolism among old non-diabetic individuals. We also explored the combined effect of birth size and midlife body mass index (BMI) on type 2 diabetes in old age. Our study comprised 1,682 Icelanders whose birth records included anthropometrical data. The same individuals had participated in the prospective population-based Reykjavik Study, where BMI was assessed at a mean age of 47 years, and in the AGES-Reykjavik Study during 2002 to 2006, where fasting glucose, insulin and HbA₁c were measured and homeostasis model assessment for the degree of insulin resistance (HOMA-IR) calculated at a mean age of 75.5 years. Type 2 diabetes was determined as having a history of diabetes, using glucose-modifying medication or fasting glucose of >7.0 mmol/l. Of the participants, 249 had prevalent type 2 diabetes in old age. Lower birth weight and body length were associated with higher fasting glucose, insulin, HOMA-IR and HbA₁c among old non-diabetic individuals. Higher birth weight and ponderal index at birth decreased the risk for type 2 diabetes in old age, odds ratio (OR), 0.61 [95 % confidence interval (CI), 0.48–0.79] and 0.96 (95 % CI, 0.92–1.00), respectively. Compared with those with high birth weight and low BMI in midlife, the odds of diabetes was almost five-fold for individuals with low birth weight and high BMI (OR, 4.93; 95 % CI, 2.14–11.37). Excessive weight gain in adulthood might be particularly detrimental to the health of old individuals with low birth weight.