This paper investigates feedback control of refrigeration cycles for high heat-flux cooling applications, where large transient heat loads may be present. We apply H∞ controller synthesis for disturbance rejection, with the evaporator heat-flux treated as the disturbance input. The controller synthesis is based on model linearization about a chosen operating point. We analyze model uncertainty due to the linearization error to ensure robustness of the closed-loop systems. We use a low-order, lumped-element nonlinear model for the vapor compression cycle. We obtain linearized systems at different operating points, and quantify system nonlinearity using the H∞ norm. Controllers synthesized for the chosen nominal systems are tested for both nominal (near the operating point) and the worst-case performance in nonlinear simulations. For systems close to critical heat-flux (CHF), it is shown that a trade-off exists between the nominal performance and robust stability. For systems far away from CHF, it is shown that the open-loop system has the optimal cooling capacity. The performance of H∞ controller for systems near CHF is validated by experiment.