The aim of the present study was to determine whether fatigue affects internal and external load variables determining power profile in cyclists. Ten cyclists performed outdoor power profile tests (lasting 1-, 5 and 20-min) on two consecutive days, subject either to a fatigued condition or not. Fatigue was induced by undertaking an effort (10-min at 95% of average power output obtained in a 20-min effort followed by 1-min maximum effort) until the power output decreased by 20% compared to the 1-min power output. Fatigued condition decreased power output (p < 0.05, 1-min: 9.0 ± 3.8%; 5-min: 5.9 ± 2.5%; 20-min: 4.1 ± 1.9%) and cadence in all test durations, without differences in torque. Lactate decreased in longer efforts when a fatigue protocol had previously been conducted (e.g., 20-min: 8.6 ± 3.0 vs. 10.9 ± 2.7, p < 0.05). Regression models (r2 ≥ 0.95, p < 0.001) indicated that a lower variation in load variables of 20-min in fatigued condition compared with the non-fatigued state resulted in a lower decrease in critical power after the fatigue protocol. The results suggest that fatigued condition on power was more evident in shorter efforts and seemed to rely more on a decrease in cadence than on torque.