The globalization of engineering activities is a recent, growing phenomenon (Harfi, Mathieu & Pfister, 2007). It is particularly noteworthy in the car industry, where the question of why and how to go about that globalization is a major challenge for firms (Calabrese, 2001; Sagiyama & Fujimoto, 2000; Miller, 1994). This paper is based on three years of cooperation with a European car manufacturer where we adopted a research-intervention methodology (Hatchuel & David, 2008; Sardas & Lefebvre, 2004). In our exploration of the implications of the creation of a new engineering unit abroad, we have mobilized two analytical models which enable us to understand the large staff turnover observed during the creation of the new unit. The "internal R&D professional group dynamics" ("dynamique de metier",) (Lefebvre, Roos & Sardas, 2002; Roos, 2006), hereafter referred to simply as the "professional group dynamics", is above all an analytical model in which three interacting sub-dynamics are distinguished: the collective dynamics of expertise (Nonaka & Takeuchi, 1995); the dynamics of roles and their interactions; and the dynamics of collective and individual identities. It is secondly a tool for structuring and management, in so far as it acts primarily on the systems of roles offered to individuals, as well as on the modalities of knowledge management. In the "global actor dynamic" model (Sardas, 1994), an individual's dynamic is seen as three partial, interlinking dynamics: a cognitive dynamic, a strategic dynamic (Crozier & Friedberg, 1977) and a subjective dynamic (Sardas, 2001). This model is designed to diagnose the encounter between an individual and an organization, from a management science point of view. Our analysis of the firm's first experiments in globalizing its engineering unit shows that the relational deficiencies resulting from actors' games led to blockage in the designers' learning dynamics and caused their dynamic of subjective investment to dwindle. As a result they either became apathetic or they resigned. We can conclude that three conditions are necessary if a designer is to learn: an activity in which new knowledge is acquired; a co-located expert collective that provides the conditions for adequate technical supervision; and the presence of related expertise. To analyse the subject in depth and to shed light on concrete decisions (how many people were recruited, how many expatriates, etc.), we developed a numeric model for simulating the process of growth and expansion of the competencies of an international engineering unit. This enabled us to assess the time-scale of this process and its sensitivity to staff turnover, which is a major risk in this type of project. Finally, by reviewing our approach and the models and reasoning applied, we examine the contingency factors of this globalization model.