<i>Background:</i> Emerging evidence indicates that hyperoxia is a risk factor for bronchopulmonary dysplasia, a common multifactorial long-term complication of prematurity. To date, the equivalence between set and delivered oxygen (O₂) in ventilated preterm infants has not been rigorously studied. <i>Objectives:</i> To test the hypothesis of systematic underestimation of O₂ delivery in extremely low birth weight (ELBW) infants during long-term ventilation. <i>Methods:</i> Actually achieved O₂ concentrations were measured and compared to the set inspired oxygen fraction (FiO₂). A total of 108 O₂ measurements were carried out during the ventilation of 54 ELBW infants: O₂-Δ error (i.e., the difference between O₂ concentrations achieved by the ventilator and set FiO₂) was the main study outcome measure. <i>Results:</i> Systematic O₂-Δ errors were found, with mean values of +9.52% (FiO₂ 0.21–0.40), +2.10 (FiO₂ 0.41–0.60), +2.86% (FiO₂ 0.61–0.80), and +0.016% (FiO₂ 0.81–1.0; p < 0.0001). Theoretical simulations from the observed data indicate that, if not corrected, systematic O2-Δ errors would lead to a non-intentional total O₂ load of 1,202.9 (FiO₂ 0.21–0.40), 252.46 (FiO₂ 0.41–0.60), 342.85 (FiO₂ 0.61–0.80), and 2 (FiO₂ 0.81–1.0) extra liters/kg body weight/100 ventilation hours. <i>Conclusions:</i> Systematic underestimation of the O₂ delivered by infant ventilators can potentially lead to surprisingly large increases in total O₂ load during long-term ventilation of ELBW infants, especially in the lower FiO₂ range (i.e., 0.21–0.40). Underestimation of true O₂ delivery can potentially lead to unrecognized high O₂ loads, and more pronounced and prolonged hyperoxia.