Altered functioning of the hypothalamic-pituitary-adrenal (HPA) axis has been demonstrated in patients with treatment-resistant depression, although studies have often conflated patients with unipolar and bipolar depression. This is problematic given that the two groups often present with opposed neurovegetative symptom patterns. The aim of this study was to test, for the first time, whether post-awakening cortisol, a highly reliable, naturalistic measure of HPA functioning, could distinguish patients with clearly defined treatment-resistant unipolar (TRUD) and bipolar depression (TRBD). A total of 37 patients with TRUD, 17 patients with TRBD, and 47 healthy controls were recruited. Areas under the curve (AUC) with respect to the ground (g) and increase (i) of post-awakening cortisol concentrations (awakening, +15, +30, +45, +60, +90 min) were measured over two days. Patients with TRUD had higher total cortisol production in the morning hours compared to controls (AUCg, p = 0.01), while they did not differ in terms of the awakening response (AUCi, p = 0.28). By contrast, subjects with TRBD had lower total cortisol when compared to controls by trend (AUCg, p = 0.07), while they did not differ in the awakening response (AUCi, p = 0.15). A direct comparison of TRUD and TRBD revealed differences in the AUCg (p = 0.003) and AUCi (p = 0.03). This finding of comparatively elevated HPA axis activity in the morning in TRUD and attenuated HPA axis activity in TRBD attests to a fundamental biological distinction between unipolar and bipolar depression. It has implications for the understanding and treatment of bipolar depression and in differentiating the two types of depression.