There is now significant evidence that initial use of the correct antibiotic saves more lives than virtually all other intensive care therapy. This means covering all possible causative organisms with the initial empirical choice. For nosocomial sepsis, broad-spectrum antibiotics must be started as soon as the relevant samples have been taken for culture, with de-escalation of therapy targeted to the causative organisms when results and susceptibilities are available. There is an international trend to use shorter antibiotic courses. Pseudomonas pneumonia probably needs a 7-10 day course. In our ICU, provided the infection source is controlled, we seldom use antibiotic courses longer than 7 days. Evaluation of the kill characteristics of antibiotics in experimental models suggests that different classes of antibiotics should have different dosing regimens. For Beta- lactam antibiotics, the kill characteristic is almost entirely related to the time that tissue and plasma levels exceed a certain threshold, with no significant post-antibiotic effect, particularly against gram-negative organisms. Kill characteristics of other antibiotics, such as aminoglycosides, relate to adequate peak concentrations and a significant post-antibiotic effect. Clinically, these kill characteristics translate into the need for appropriate doses of the various antibiotics in patients with sepsis. We have shown that some patients with normal serum creatinine levels have very high creatinine clearance rates; in ICU patients with sepsis, blood pressure and tissue perfusion are maintained with large fluid loads and inotropic agents, thereby raising creatinine clearance. High clearances produce low trough concentrations of antibiotic, with important implications for underdosing and the development of antibiotic resistance. The new paradigm for treating sepsis, particularly nosocomial sepsis, is: get it right the first time, hit hard up front, and use large doses of broad-spectrum antibiotics for a short period.