The aims were to optimize a nuclear magnetic resonance (NMR)-based assay for quantifying ionized or free magnesium and investigate its association with type 2 diabetes (T2D). A high-throughput, ionized magnesium assay was optimized and evaluated. Plasma magnesium was quantified, and associations with T2D were ascertained in Insulin Resistance Atherosclerosis Study (IRAS) participants. Coefficients of variation for the ionized magnesium assay ranged from 0.7–1.5% for intra-assay and 4.2–4.7% for inter-assay precision. In IRAS (n = 1342), ionized magnesium was significantly lower in subjects with prediabetes and T2D than in normoglycemic subjects, and lower in participants with T2D than those with prediabetes (p < 0.0001). Cross-sectional regression analyses revealed that magnesium was associated with T2D at baseline in models adjusted for multiple clinical risk factors (p = 0.032). This association appeared to be modified by sex, in such a way that the associations were present in women (OR = 0.54 (95% CI 0.37–0.79), p = 0.0015) and not in men (OR = 0.98 (95% CI 0.71–1.35), p = 0.90). Longitudinal regression analyses revealed an inverse association between magnesium and future T2D in the total population (p = 0.035) that was attenuated by LP-IR (p = 0.22). No interactions were detected between magnesium and age, race, BMI, glucose, insulin, triglycerides, or LPIR for the prospective association with future T2D. However, a significant interaction between magnesium and sex was present, now with a trend for an association in men (OR = 0.75 (95% CI 0.55–1.02), p = 0.065 and absence of an association in women (OR = 1.01 (0.76–1.33), p = 0.97). Conclusions: lower ionized magnesium, as measured by an NMR-based assay optimized for accuracy and precision, was associated cross-sectionally with T2D at baseline and longitudinally with incident T2D in IRAS.