Retirement policy is often predicated on the belief that more saving is always better, at least at the margin. This belief is used to justify the increasingly widespread practice of automatically enrolling workers in employer-sponsored defined contribution plans. However, the conclusion that individuals do not save optimally for retirement requires a benchmark for optimal behavior. A reasonable benchmark that is often used in the academic literature is the life-cycle model, in which rational individuals allocate resources over their lifetimes with the aim of avoiding sharp changes in their standard of living. We argue that, under realistic assumptions, the life-cycle model implies that most young people should not save for retirement. First, high-income workers tend to experience wage growth over their careers. For these workers, maintaining as steady a standard of living as possible therefore requires spending all income while young and only starting to save for retirement during middle age. Second, low-income workers, whose wage profiles tend to be flatter, receive high Social Security replacement rates, making optimal saving rates very low. Finally, for all workers, low real interest rates make a front-loaded lifetime spending profile optimal. We show that the welfare costs of automatically enrolling younger workers in defined contribution plans — if they are passive savers who do not opt out immediately — can be substantial, even with employer matching.