This paper extends the analysis of the long-run pricing and capacity d
ecision problem for shared computer services by Dewan and Mendelson (1
990) and makes two further contributions. First, we show that simple m
arginal capacity cost pricing is often optimal in the absence of priva
te user information, and it outperforms cost recovery and profit cente
r pricing methods. Second, we provide insights into the implications o
f declining computing costs on the tradeoff between capacity costs and
user time. In equilibrium, expected user delay costs are bounded by c
apacity costs due to the substitution of cheaper information processin
g capacity for valuable user time.