We introduce stochastic R&D in the Hotelling model and show that if the technical risk is sufficiently high, all firms focus on the most valuable market segment. We then endogenize technical risk by allowing firms to choose between a safe and a risky R&D technology. Firms either both target the most attractive market with at least one firm using the risky technology or they choose different niche projects and both apply the safe technology. R&D spillovers lead to more differentiated R&D projects and patent protection to less. Project coordination within an RJV implies more differentiation, and may be welfare-improving.