In emerging economies, infrastructure projects are in full swing. There is a wealth of replicable experience for exploitation. Simultaneously, more technologies and methodologies require further exploration. This makes fostering ambidextrous innovation strategies (i.e., the tradeoff between exploitative and exploratory innovation strategies) a common and vital practical issue. Large infrastructure projects are unique one-off endeavors but have somewhat repetitive and persistent characteristics. It is a particular “intermediate” form between temporary projects and permanent organizations. Previous research on fostering ambidextrous innovation strategies cannot simply be replicated in large infrastructure projects. To address this issue, this article investigates the relationship between team heterogeneity and ambidextrous innovation strategies and also the role of team learning and identification in large infrastructure projects. Data were collected from 269 responses from 31 large infrastructure project delivery teams in China. The findings show that team heterogeneity has a positive linear effect on exploratory and ambidextrous innovation strategies and an inverted U-shaped effect on exploitative innovation strategies; team heterogeneity can better foster ambidextrous innovation strategies through improving team learning; the moderating role of team identification in the overall mechanism differs from the usual assumptions in permanent organizations. Overall, this article extends the existing ambidexterity research in the “intermediate” form between temporary projects and permanent organizations. It provides insights and guidance on fostering ambidextrous innovation strategies in large infrastructure projects.