Our scarcest resource is time. Humans and other animals allocate their time on a vast number of different activities – from reading a paper, pursuing a degree, or foraging for food. Optimal, reward-maximizing, behavioral strategies must balance expected future gains with irrecoverable time investment. However, humans and other animals often succumb to cognitive biases that lead to suboptimal payoffs such as considering sunk costs when investing time to obtain future payoffs. Here we argue that humans, rats, and mice invest time according to optimal predictions given available evidence and subjective time preferences. First, we argue that previous data showing apparent sunk cost sensitivity (1) can be explained by an attrition bias when analyzing variable choice behavior of rational agents. Second, we behaviorally tested the optimality of time investments in humans, rats, and mice. Across trials we varied subjects’ uncertainty in obtaining rewards, enabling us to evaluate whether subjects invested time in proportion to their confidence about successful outcomes. Human or animal subjects committed to one of two choice alternatives with ambiguous sensory evidence, which determined the subjects’ confidence in making a correct choice. After making a choice, subjects invested time in their decision by waiting for randomly delayed reward for correct choices. We developed a statistical approach to non-parametrically estimate the optimal (reward-maximizing) time investment given the evidence and their overall time preference. Crucially, this approach allowed us to isolate the contribution of confidence to investment behavior irrespective of individuals’ subjective assessment of the costs and benefits of waiting. We found that all species invested time close to optimal model predictions, i.e., with low confidence noise, demonstrating that all species appropriately use sensory evidence to adjust investments. Our analyses reveal that humans, rats, and mice can be efficient economic agents, who appropriately allocate investment decisions in uncertain environments.