With the goal of increasing competition in the telecommunications sector, regulatory authorities around the world have adopted cost-based prices for network interconnection and access services. The process of introducing competition has not been easy and many issues have arisen in recent years related to which facilities should be made available by the incumbent carriers, and on what terms and conditions. One of these issues, not yet addressed by the regulatory authorities, is the consideration of the value of the option to invest when calculating cost-based prices of regulated services. The pricing formula currently used by the regulators ignores the value of the option to productively invest at some time in the future. This thesis proposes a model and methodology for valuing the option to delay network investment decisions and calculating costbased prices of regulated telecommunications services, taking into account the demand and technological uncertainties in telecommunications networks. In a typical cost study, the telecommunications network is represented by a list of network elements specifically dimensioned to meet the forecast demand for all telecommunications services, where each network element is an identifiable part of the network infrastructure (e.g., the local loop), for which it can be assigned a single cost driver. Different network elements are subject to different demand and technological uncertainties. For example, switches and transmission equipment are subject to faster technological substitution than local loop and transmission facilities. In this thesis, I calculate the option value multiples for the decisions to invest in three main network elements, each representing a different part of the Brazilian fixed telecommunications network (subject to different technological and demand uncertainties), and estimate the impact of these option value multiples on the average unitary cost of each network element. This thesis innovates in several aspects. First, because different network elements are subject to different demand and technological uncertainties, a markup factor is calculated for each main network element. Second, the value of the option to invest in each network element is modeled as a function of two stochastic variables: the element s total variable profit and the cost of new investment in the element. Third, technological uncertainty is modeled using two complementary approaches: technology obsolescence of used equipment and technology evolution of new equipment. Fourth, the value of future replacement options is considered, allowing for the resizing of network capacity as equipment elements are replaced. The demand and technological uncertainties associated with each network element are modeled through the use of three stochastic processes: the flow of total variable profit (geometric Brownian motion), the depreciation of used asset (Poisson decay process), and the cost of modern equivalent asset (geometric Brownian motion). They all fit together into a neat and simple model that calculates the option value multiple for each network element. A constant riskfree interest rate has been assumed to derive the stochastic differential equations that the real option values must satisfy, although interest rate uncertainty has been investigated and interest rate volatilities have been calculated for different maturities/terms. This thesis proposes a method for constructing the Brazilian inflation coupon curve using a combination of traditional nonlinear optimization algorithm and a genetic algorithm specifically developed for that purpose. There has been a good deal of debate about which markup factor (if any) should be applied to the investment cost component of a network investment decision in order to reflect the value of the killed option. Some authors say the real option value is negligible and should be ignored, as in Pelcovits (1999), while others calculate markup values that are quite significant, as in Hausman (1999) and Pindyck (2005). This thesis shows that the markup values can be negligible for some network elements and quite significant for others. After applying the markup factors, network costs should be assigned to services on the basis of how much each service uses each network element. The impact of the option value multiples on the cost-based prices of regulated telecommunications services will depend on how much each service uses each network element. The proposed model and methodology address a problem left unsolved over the past 14 years and might become a landmark in regulatory cost modeling and cost-based access pricing. Although I focus on telecommunications, model and methodology can be adapted for application to other network industries as well.