Fetal macrosomia may occur even in adequately controlled diabetic mothers. This may reflect the problem of using maternal glycemia as an indicator of fetal glycemia, because the placenta interposed between both compartments has its own glucose metabolism. Here, we propose a model by which the placenta protects the fetus at moderate levels of maternal hyperglycemia. One characteristic feature of the human placenta in diabetes is the increased deposition of glycogen. Neither hyperglycemia nor hyperinsulinemia increase the glycogen content in the trophoblast. Since the glycogen increments in diabetes are predominantly located around fetoplacental vessels, it is tempting to assume a fetal origin of glucose making up the glycogen deposits. In fact, glucose can be transported back from the fetus into the placenta and this reflux is increased in diabetes. Therefore, in conditions of fetal glucose levels exceeding the demand for sustaining fetal growth and metabolism, glucose can be stored in the liver and other fetal tissues. Once these stores are saturated, glucose is extracted from the fetal circulation by the glucose transporters GLUT1 and GLUT3 on cells surrounding the fetoplacental vasculature and stored therein, again in the form of glycogen. These processes might be under the control of fetal insulin, because insulin injected into the fetal circulation increases placental glycogen stores. Fetal macrosomia would then occur only when fetal hyperglycemia exceeds the placental capacity to store excess fetal glucose. Thus, the placental failure to protect the fetus would cause the 'unexplained' phenotypic changes occasionally found in fetuses born to well-controlled diabetic women.