Cooling of gas turbine blades is critical to long term durability. Accurate prediction of blade metal temperature is a key component in the design of the cooling system. In this design space, spatial distribution of heat transfer coefficients plays a significant role. Large-Eddy Simulation (LES) has been shown to be a robust method for predicting heat transfer. Because of the high computational cost of LES as Reynolds number (Re) increases, most investigations have been performed at low Re of O(10⁴). In this paper, a two-pass duct with a 180° turn is simulated at Re = 100,000 for a stationary and a rotating duct at Ro = 0.2 and Bo = 0.5. The predicted mean and turbulent statistics compare well with experiments in the highly turbulent flow. Rotation-induced secondary flows have a large effect on heat transfer in the first pass. In the second pass, high turbulence intensities exiting the bend dominate heat transfer. Turbulent intensities are highest with the inclusion of centrifugal buoyancy and increase heat transfer. Centrifugal buoyancy increases the duct averaged heat transfer by 10% over a stationary duct while also reducing friction by 10% due to centrifugal pumping.