What produces a hurricane or a typhoon? Despite years of effort by many meteorologists, the question has not yet been completely answered. To prepare a general background for understanding this problem, a brief survey is given of the structure and dynamics of fully developed tropical cyclones. Emphasis is given to the warm‐core thermal structure, which has a significant role in driving the three‐dimensional circulation, and to kinetic energy of mature tropical cyclones. Examination of the thermodynamic energy balance shows that the thermal structure is maintained by release of latent heat of condensation of water vapor. The importance of the processes taking place in the surface boundary layer is also discussed. Despite the apparent simple thermal structure of the mature cyclone, its formation process has turned out to be much more complex than it was once thought to be. A simple buoyant convective hypothesis will not adequately explain the observed fact; thus other factors become significant, especially large‐scale flow patterns such as the equatorial shear zone, waves in the easterlies, and high‐tropospheric migratory vortices. An analysis of the transient formative stage in which an originally cold‐core wave disturbance is transformed into an incipient warm‐core vortex is presented. According to this analysis, three characteristic stages in the whole process of cyclone formation are defined. Next we proceed to a summary of the progress of theoretical and numerical studies relating to tropical cyclone formation. The thermal (buoyant) instability theory, based on the negative vertical gradient of equivalent potential temperature, allows cumulus‐scale convection but does not directly explain tropical cyclone formation. This has been confirmed by several numerical models using high‐speed electronic computers. In reality, the tropical cyclone seems to possess two different convection characteristics, i.e., the macroscopic cyclone itself and cumulus clouds embedded within. Thus we need to advance a theory which explains the growth of the macroscopic cyclone by parameterizing the role of cumulus clouds. Two possible dynamical frameworks are considered. These are the ‘balanced’ and ‘unbalanced’ baroclinic vortex motions with a radial‐vertical circulation. The latter theory involves the inertial instability mechanism along the vertically inclined isentropic surfaces. On the basis of facts and theoretical deductions, an attempt is made to develop a plausible model of tropical cyclone formation. The mechanisms considered important in each of the three postulated stages of formation are discussed. These are: (1) the pre‐existing large‐scale vertical motion associated with waves in the easterlies; (2) the organization of cumulus convection by the large‐scale vertical motion and the role of the cumulus convection in heat liberation and vertical transport; and (3) the formation of a warm‐core thermal field and the development of an incipient baroclinic vortex into a mature cyclone. Finally, the potential importance of several other studies, especially the study of mutual relationship between cumulus convection and the tropical cyclone, is mentioned. The need for expansion of observation networks in the tropics is emphasized in relation to several fundamental problems of tropical circulation.