Obesity is a risk factor for the appearance of type 2 diabetes: this notion, largely based on clinical experience, dictates the essential principles of the treatment of type 2 diabetes. At odds with this conventional wisdom, Elliot Danforth Jr. has recently proposed that" too few adipocytes predisposes to type 2 diabetes". A further thought on this controversy is discussed herein. In the context of the" orthodox" view which links obesity and diabetes, and its relation to the controversy, we analyse on one hand the effects of thiazolidine-diones on insulin sensitivity and on adipogenesis and, on the other hand, those of extreme situations represented by lipoatrophic diabetes and morbid obesity. This analysis shows that fat tissue indeed favors the appearance of diabetes but is also able to be anti-diabetogenic, and a dynamic solution of this paradox is put forward. We propose a dual evolutionist hypothesis leading to the selection of an intermediary adipogenic genotype based on the limitation of both insulin secretion and adipogenic potential, which would explain both the necessity of the existence of adipose tissue in man and the limitations of its development.