We theoretically predict a new phenomenon, namely, that a solid−solid phase transformation (PT) with a large transformation strain can occur via internal stress-induced virtual melting along the interface at temperatures significantly (more than 100 K) below the melting temperature. We show that the energy of elastic stresses, induced by transformation strain, increases the driving force for melting and reduces the melting temperature. Immediately after melting, stresses relax and the unstable melt solidifies. Fast solidification in a thin layer leads to nanoscale cracking which does not affect the thermodynamics or kinetics of the solid−solid transformation. Thus, virtual melting represents a new mechanism of solid−solid PT, stress relaxation, and loss of coherence at a moving solid−solid interface. It also removes the athermal interface friction and deletes the thermomechanical memory of preceding cycles of the direct-reverse transformation. It is also found that nonhydrostatic compressive internal stresses promote melting in contrast to hydrostatic pressure. Sixteen theoretical predictions are in qualitative and quantitative agreement with experiments conducted on the PTs in the energetic crystal HMX. In particular, (a) the energy of internal stresses is sufficient to reduce the melting temperature from 551 to 430 K for the δ phase during the β → δ PT and from 520 to 400 K for the β phase during the δ → β PT; (b) predicted activation energies for direct and reverse PTs coincide with corresponding melting energies of the β and δ phases and with the experimental values; (c) the temperature dependence of the rate constant is determined by the heat of fusion, for both direct and reverse PTs; results b and c are obtained both for overall kinetics and for interface propagation; (d) considerable nanocracking, homogeneously distributed in the transformed material, accompanies the PT, as predicted by theory; (e) the nanocracking does not change the PT thermodynamics or kinetics appreciably for the first and the second PT β ↔ δ cycles, as predicted by theory; (f) β ↔ δ PTs start at a very small driving force (in contrast to all known solid−solid transformations with large transformation strain), that is, elastic energy and athermal interface friction must be negligible; (g) β → α and α → β PTs, which are thermodynamically possible in the temperature range 382.4 < θ < 430 K and below 382.4 K, respectively, do not occur.