Physics imposes fundamental constraints on power amplifier (PA) efficiency. PA is unambiguously thermodynamic system that generates waste heat. One physical constraint concerns heat dissipation which limits the size and efficiency of PA. The irreversible heat loss generated during signal amplification process reduces the efficiency of a PA. To date, very few methods are available for the design of PAs; many of the design processes have been empirical. The theoretical work in this thesis attempts to describe an analytical method for analysing PA's efficiency using the underlying principles of physics. The analysis starts from the second law of thermodynamics and the concept of entropy. Entropy upper bounds of a physical system is determined from the black hole thermodynamics and Stefan-Boltzmann laws of thermal radiation. The result helps to formulate an argument to the scaling of electronic devices. 1) The entropy of a physical system is ultimately bounded in proportional to its surface area. 2) Device scaling is a proper way to improve energy conversion efficiency in the sense of reducing heat dissipation. Quantum mechanics and thermodynamics impose limits on scaling. The limits of scaling from the constraint of heat removal has been developed. A figure of merit that relates switching speed and minimum chip area is proposed. A novel one-dimensional method for the fast computation of the thermal characteristics of GaAs HBTs PA has been developed. The most effective form factors of heat removal spreader structure are investigated and given. Finally, a nanoscale model based on drift-diffusive and quantum transport approach with non-equilibrium Green's function method, for the computation of the MOSFET switching loss, is proposed and developed. The model is valid for MOSFET with channel lengths above the electron de Broglie wavelength (~10 nm). The study empowers MOSFET switching loss to be determined from the contact materials' wavefunctions, channel length, source and drain doping concentration, DC bias, switching frequency and ambient temperature.