Over the past two decades, most guidelines have advocated early dialysis initiation on the basis of studies showing improved survival in patients starting dialysis early. These recommendations led to an increase in the proportion of patients initiating dialysis with an estimated glomerular filtration rate (eGFR) >10 ml/min/1.73 m2, from 20% in 1996 to 52% in 2008. During this period, patients starting dialysis with an eGFR ≥15 ml/min/1.73 m2 increased from 4% to 17%. However, recent studies have failed to substantiate a benefit of early dialysis initiation and some data have suggested worse outcomes in patients starting dialysis with a higher eGFR. Several reasons for this seemingly paradoxical observation have been suggested, including the fact that patients requiring early dialysis are likely to have more severe symptoms and comorbidities, leading to confounding by indication, as well as biological mechanisms that causally relate early dialysis therapy to adverse outcomes. Dialysis reinitiation in patients with a failing renal allograft encounters similar problems. However, unique factors associated with a failed allograft means that the optimal timing of dialysis initiation in failed transplant patients might differ from that in transplant-naive patients. In this Review, we will discuss studies of dialysis initiation and compare risks and benefits of early versus late dialysis therapy.