Competitive and recreational cyclists use stiff-soled shoes that firmly attach to ‘clipless’ pedals. When compared to very flexible running shoes, very stiff cycling shoes with clipless pedals enhance power output during sprint cycling. However, a recent study showed no difference in power output or sprint performance between commercially available cycling shoes that span a range of longitudinal bending stiffnesses (∼200 to 500 N·m/rad). Thus, we sought to identify the likely range of sole stiffnesses below 200 N·m/rad over which reduced cycling shoe sole stiffness begins to decrease maximal power output. We measured the mechanical power outputs of 25 road cyclists during maximal sprints wearing shoes with identical uppers but with three different sole stiffnesses. Each participant completed nine 50 m sprints (three trials for each shoe) on a road with a steady, uphill grade of 9.1%. The three shoe sole conditions were: injected nylon (longitudinal bending stiffness 194 N·m/rad), moderate stiffness thermoplastic polyurethane (medium TPU) (43 N·m/rad), and low stiffness TPU (soft TPU) (9 N·m/rad) all ridden with the same clipless pedals. Stiffness was quantified using a simple testing apparatus. Power output decreased below the sole stiffness of ∼200 N·m/rad but only moderately. Maximal 1 s crank power (Pmax1) decreased −3.1% (ES = −0.59, p = 0.020) from Nylon to medium TPU and then further decreased −2.4% (ES 16 = −0.50, p = 0.054) from medium TPU to soft TPU. Interpolating our results suggests just a ∼1% loss in Pmax1 at a sole stiffness of 100 N·m/rad. Within reasonable limits, cycling shoe sole stiffness has only a small effect on power output.