Firms jointly sell durable primary hardware and complementary software in many industries, but it remains unclear how they can coordinate the two products to conduct intertemporal price discrimination (IPD). In addition to the harvesting (i.e., pricecutting over time) incentive in the traditional single-product IPD literature, firms selling complementary products have another investing (i.e., price-raising over time) incentive to penetrate the market early and earn from subsequent complementary product sales. We study the optimal pricing strategy and profitability of joint IPD in the context of e-readers and e-books. A forward-looking monopolist firm makes dynamic pricing decisions for both e-readers and e-books, and forward-looking consumers make dynamic e-reader and book purchase decisions anticipating future price changes. We structurally estimate the demand system using individual-level transaction data of e-readers and books from 2008 to 2012. The estimation reveals two consumer types, "avid readers" and "general readers," who self-select into buying e-readers based on their unobserved heterogeneous tastes for books. Given the demand estimates, we numerically solve for the optimal joint IPD policy and find that harvesting on e-readers and investing in e-books is optimal. In particular, the optimal pricing policies are functions of the composition of consumer types. The monopolist should harvest on e-readers and invest in e-books for avid readers, while it should invest in e-readers and harvest on e-books for general readers. We explain this difference by comparing the relative demand elasticities between e-readers and e-books for avid and general readers. The joint IPD policy provides a better screening device for the monopolist because different price path combinations can induce different consumer types to purchase. It also limits consumers' ability to arbitrage on price changes over time. The profitability of the joint IPD policy depends on the composition of heterogeneous consumers in the initial market. We find that conducting IPD on both products may hurt firm profits if the fraction of avid readers is too low.