The giant gas planets have hot convective interiors, and therefore a common assumption is that these deep atmospheres are close to a barotropic state. Here we show using a new anelastic general circulation model that baroclinic vorticity contributions are not negligible, and drive the system away from an isentropic and therefore barotropic state. The motion is still aligned with the direction of the axis of rotation as in a barotropic rotating fluid, but the wind structure has a vertical shear with stronger winds in the atmosphere than in the interior. This shear is associated with baroclinic compressibility effects. Most previous convection models of giant planets have used the Boussinesq approximation, which assumes the density is constant in depth; however, Jupiter's actual density varies by four orders of magnitude through its deep molecular envelope. We therefore developed a new general circulation model (based on the MITgcm) that is anelastic and thereby incorporates this density variation. The model's geometry is a full 3D sphere down to a small inner core. It is nonhydrostatic, uses an equation of state suitable for hydrogen-helium mixtures (SCVH), and is driven by an internal heating profile. We demonstrate the effect of compressibility by comparing anelastic and Boussinesq cases. The simulations develop a mean state that is geostrophic and hydrostatic including the often neglected, but significant, vertical Coriolis contribution. This leads to modification of the standard thermal wind relation for a deep compressible atmosphere. The interior flow organizes in large cyclonically rotating columnar eddies parallel to the rotation axis, which drive upgradient angular momentum eddy fluxes, generating the observed equatorial superrotation. Heat fluxes align with the axis of rotation, and provide a mechanism for the transport of heat poleward, which can cause the observed flat meridional emission. We address the issue of over-forcing which is common in such convection models and analyze the dependence of our results on this; showing that the vertical wind structure is not very sensitive to the Rayleigh number. We also study the effect of rotation, showing how the transition from a rapidly to a slowly rotating system affects the dynamics.