Innovation is a central yet ambiguous goal in the European Union’s digital agenda. Within the AI Act, innovation is positioned as both an economic imperative and a normative principle. Globally, however, innovation is often equated with unbounded growth and the belief that technological progress is inherently beneficial, sometimes even magical – an outlook rooted in Silicon Valley techno-optimism. Scholarship has examined how these imaginaries shape AI discourse primarily in the US context; less attention has been paid to European AI governance. Drawing on qualitative interviews with eleven EU digital policy experts, this study explores how innovation is articulated in the EU AI Act. The findings demonstrate competing innovation narratives that (partly) challenge techno-optimist narratives, converging around what I call institutional optimism: a belief that locates transformative potential not in technology itself, but in institutional intervention as the necessary mechanism for realising its benefits and mitigating its harms.