Type 2 diabetes (T2D) is characterized by increased fracture risk despite higher BMD and reduced bone turnover. BMD underestimates fracture risk in T2D, but the predictive role of bone turnover markers (BTMs) on fracture risk in T2D has not been explored. Thus, we sought to determine whether BTMs predict incident fractures in subjects with T2D. For this case‐cohort study, we used data from the Health, Aging, and Body Composition (Health ABC) Study of well‐functioning older adults, aged 70 to 79 years at baseline (April 1997–June 1998). The case‐cohort sample consisted of (i) the cases, composed of all 223 participants who experienced incident fractures of the hip, clinical spine, or distal forearm within the first 9 years of study follow‐up;​ and (ii) the subcohort of 508 randomly sampled participants from three strata at baseline (T2D, prediabetes, and normoglycemia) from the entire Health ABC cohort. A total of 690 subjects (223 cases, of whom 41 were in the subcohort) were included in analyses. BTMs (C‐terminal telopeptide of type I collagen [CTX], osteocalcin [OC], and procollagen type 1 N‐terminal propeptide [P1NP]) were measured in archived baseline serum. Cox regression with robust variance estimation was used to estimate the adjusted hazard ratio (HR) for fracture per 20% increase in BTMs. In nondiabetes (prediabetes plus normoglycemia), fracture risk was increased with higher CTX (HR 1.10;​ 95% confidence interval [CI], 1.01 to 1.20 for each 20% increase in CTX). Risk was not increased in T2D (HR 0.92;​ 95% CI, 0.81 to 1.04;​ p for interaction .045). Similarly, both OC and P1NP were associated with higher risk of fracture in nondiabetes, but not in T2D, with p for interaction of .078 and .109, respectively. In conclusion, BTMs did not predict incident fracture risk in T2D but were modestly associated with fracture risk in nondiabetes.