The Gulf’s artificial intelligence (AI) boom is commonly framed as a project of economic diversification, yet it also raises questions of governance, security and technological dependence in states whose constitutional orders explicitly incorporate Islamic identity or principles. This article comparatively examines the development and governance of AI in Saudi Arabia, the United Arab Emirates (UAE) and Qatar, focusing on the relationship between hydrocarbon-derived state capacity, technological localisation and external dependence. Drawing on a structured analysis of public documents issued between 2017 and August 2026, including national AI strategies, ethics and cybersecurity frameworks, sovereign investment initiatives, corporate partnerships and export-control decisions, the study traces how financial and energy resources are being translated into computational infrastructure and domestic AI capability. It also examines how Islamic constitutional and normative commitments interact with the increasingly transnational language of AI governance. The analysis identifies distinct national trajectories within a common structural pattern. All three states have expanded domestic infrastructure, institutional capacity and bargaining leverage, yet continue to depend substantially on foreign semiconductors, cloud services, frontier AI models and externally governed technology supply chains. Islamic principles provide an important constitutional and normative foundation for ideas of public interest, responsibility, dignity and social order, but operational AI governance increasingly combines these commitments with internationally circulating ethical and regulatory frameworks. The article argues that Gulf AI development is producing neither straightforward technological sovereignty nor simple technological dependence. Instead, it reflects a form of negotiated technological autonomy in which localisation increases strategic capacity while simultaneously reorganising, rather than eliminating, dependence on external technology networks.