In this paper, we investigate the asymptotic behaviour of stochastic pantograph delay evolution equations driven by a tempered fractional Brownian motion (tfBm) with Hurst parameter H > 1/2. First of all, the global existence, uniqueness and mean square stability with general decay rate of mild solutions are established. In particular, we would like to point out that our analysis is not necessary to construct Lyapunov functions, but we deal directly with stability via the Banach fixed point theorem, the fractional power of operators and the semigroup theory. It is worth emphasizing that a novel estimate of stochastic integrals with respect to tfBm is presented, which greatly contributes to the stability analyses. Then after extending the factorization formula to the tfBm case, we construct the nontrivial equilibrium solution, defined for t ∈ R, by means of an approximation technique and a convergence analysis. Moreover, we analyze the Hölder regularity in time and general stability (including both polynomial and logarithmic stability) of the nontrivial equilibrium solution in the sense of mean square. As an example of application, the reaction diffusion neural network system with pantograph delay is considered, and the nontrivial equilibrium solution and general stability of the system are proved under the Lipschitz assumption.