We discuss the status, trends and long-term ambitions of Computational Fluid Dynamics (CFD) when applied away from the design point or concept, and therefore in the historically weak areas of CFD. This includes both undesirable flight conditions, such as stall, and undesirable products of the flow, such as noise. All pose the great challenge of turbulence, and accuracy is as dependent on the ideas behind the turbulence treatment as it is on computing power. A measured shift from Reynolds-averaged representations to large-eddy simulations will take place. Empiricism, both turbulence and engineering related, will recede only step by step over many years. Yet, CFD will make full use of every increase in computer power of this century. Increasingly, CFD will compete with flight tests, not just with wind tunnels, and will be validated by flight tests. Integration with other disciplines will allow us to predict crucial phenomena such as flutter, sonic fatigue, and pilot-induced oscillations. A gap will remain at any time between the phenomena amenable to a ‘grand challenge’ calculation and those amenable to a fully CFD-based design process, because certification involves thousands of conditions. The highest demand currently is in community and cabin noise for which industrial-accuracy methods are non-existent. On the other hand, gratifying progress has occurred in the area of stall and spin.