Conventional analysis often assumes that there are only two theoretical options for calculating a reasonable royalty in patent disputes: a “pure ex ante” approach, under which a court reconstructs the hypothetical bargain the parties would have struck prior to infringement, based on the information available to them at that time; and a “pure ex post” approach, under which the court considers the bargain the parties might have reached as of some later date such as the date of judgment. The first approach avoids patent holdup — basing the royalty partly on the infringer's sunk costs — but cannot easily explain other longstanding features of how royalties are calculated, and can lead to awards that reflect the parties’ erroneous ex ante expectations. By contrast, the pure ex post approach uses more accurate information about the invention’s actual value, but it also enables the patentee to capture some of the patent’s ex post holdup value. In this Article, we show that a “contingent ex ante” framework, under which the court reconstructs the bargain the parties would have reached ex ante, based on all relevant information that is available ex post, is superior to both of the conventional approaches. More specifically, our framework enables courts to base the royalty on the most accurate information available of patent value while avoiding the holdup risk arising from the pure ex post approach. We analyze how courts can apply our approach in various settings, including cases involving SEPs, sequential infringement, regulatory uncertainty, and unexpected exogenous events.