: ICU patients with haematological malignancy have an increased risk of bleeding. Recently, global haemostatic methods such as thromboelastography (TEG) have gained impact in evaluating coagulation. The aim of this study was to observe whether TEG could predict bleeding in haematological ICU patients with severe sepsis. Post-hoc single-centre analysis of patients with haematological malignancy included in the Scandinavian Starch for Severe Sepsis/Septic Shock (6S) trial. Clinical characteristics, TEG measurements and details regarding bleeding complications were retrieved from the 6S database. The association between TEG and bleeding were analysed by Cox regression and receiver operating characteristic curves. A total of 202 patients with severe sepsis were admitted to the ICU of Rigshospitalet, Copenhagen and included in the 6S trial. Forty-one had haematological malignancy and were analysed in the current study. During ICU stay, 20 patients (49%) had bleeding complications and 13 (32%) patients bled within the first 5 ICU days. We observed no associations between TEG and subsequent bleeding in Cox regression models. TEG variables at baseline had low predictive value for bleeding. Baseline TEG variables did not add value in identifying patients with high risk of bleeding in ICU patients with haematological malignancy and severe sepsis.