Application and development of muscle powered cardiac assist devices is limited by the ability to predict the sustainable power output of in situ conditioned muscle under the expected loading conditions and geometrical constraints. Empirical definition of the sustained power limits and representative models of the bounding conditions where continuous power can be obtained are needed for device design and optimization. The latissimus dorsi muscles of four goats were chronically conditioned for 11 weeks with an implanted myostimulator. The ability to sustain power under isotonic conditions was evaluated across a range of contraction durations (100-600 ms) and rates (10-120 contractions/min). Muscles were characterized both biomechanically and myothermically to develop and evaluate three increasingly complex empirically-based models of metabolic utilization per contraction based on (1) the duty cycle, (2) a linear function of activation time, and (3) a multivariate-derived function of contraction duration, muscle load, and shortening distance. A clearly defined boundary for sustainable stimulation conditions was observed and was best represented by the linear metabolic model. These data provide both an empirical measure of chronically sustainable muscle power and predictive metabolic models that may be used to optimize the power harnessed for skeletal muscle actuated devices.