This paper asks whether the rapid innovation in Central Processing Units (CPU) results in inefficient elimination of basic Personal Computer (PC) configurations. I estimate a model in which PC makers choose first which CPU options to offer with their products, and then set prices. I contribute to the literature by analyzing a game in which firms make multiple discrete product choices. This requires relaxing point-identifying assumptions, allowing for a large product space, and tackling sample selection problems. I find that the demand for PCs is highly segmented. Using the estimated model in counterfactual analysis, I find that Intel’s introduction of its Pentium M chip contributed significantly to the growth of the mobile segment of the PC market, and to total consumer surplus, while crowding out older technologies. The scope for inefficient product elimination appears to be very limited: the upper bound on the welfare loss appears modest, while the lower bound suggests no welfare loss. I also find that the lion’s share of the short-run effect of innovation is enjoyed by the 20 % least price-sensitive consumers. Important questions regarding complementarities in innovative activities and their associated long-term benefits are left for future research. ∗ I am indebted to my advisors, Steven Berry and Philip Haile, for their continued advice and encouragement. I am also especially grateful to Donald Andrews for his advice. I have benefited greatly from discussions with Eduardo Faingold, Joshua Lustig, fellow graduate students at Yale, and numerous seminar participants. All errors are mine. I am grateful to IDC and to Mr. Steven Clough for making data available. Financial support from the Carl Arvid Anderson Prize Fellowship of the Cowles Foundation is gratefully acknowledged.