It's not about it being just energy density, but rather, as chronon mentioned, about the pressure. Specifically, this relationship determines how quickly the energy density dilutes. If an energy density dilutes slowly, it causes an accelerated expansion. This can be trivially seen from the Friedmann equation for a flat universe:
H^2 = H_0^2\frac{\rho}{\rho_0}
...where \rho is the energy density of whatever stuff there is in the universe. The Hubble parameter, H is shorthand for \dot{a}/a, with the dot denoting a derivative with respect to time, and a being the scale factor. If the energy density is a constant (thus \rho = \rho_0), as with the cosmological constant, then we have a simple differential equation:
\frac{1}{a}\frac{da}{dt} = H_0
\frac{da}{dt} = H_0 a
Since on the left hand side we have a derivative with respect to a, and the right hand side is proportional to a, we simply have an exponential:
a(t) = a(0)e^{H_0 t}
...which is an equation for accelerated expansion.
Well, it's certainly wrong in the context of a cosmological constant. I mean, sure, if your total energy density of the universe was negative, then you would have a very large negative curvature, which would cause everything to fly apart. But since inflation forces the universe to be nearly flat, this basically can't happen, so what would happen with a negative cosmological constant (or other, similar energy density) is that you'd have a large matter density, and it would just collapse right back in upon itself.
This are different when that gas is not surrounded by vacuum, but permeates all of space.