The rapid adoption of public cloud, hybrid cloud, multi-cloud, edge computing, containers, microservices, and software-defined infrastructure has transformed modern enterprise information systems while creating increasingly complex cybersecurity challenges. Mission-critical applications now operate across dynamic and geographically distributed environments in which conventional perimeter-based security mechanisms are insufficient to address sophisticated attacks, identity compromise, cloud misconfiguration, lateral movement, and previously unknown threats. Artificial intelligence (AI), machine learning (ML), deep learning (DL), behavioral analytics, threat intelligence, and automated security orchestration provide new capabilities for detecting abnormal activities and improving cyber resilience. This paper examines an AI-driven threat detection and resilient cloud security architecture designed for modern enterprise systems. The proposed architecture integrates continuous telemetry collection, intelligent anomaly detection, identity-centric zero-trust controls, threat intelligence, risk-based decision-making, automated response, and recovery mechanisms. A qualitative research methodology based on systematic literature analysis and conceptual architectural evaluation is adopted to investigate the effectiveness, scalability, adaptability, and trustworthiness of intelligent cloud defense. Recent research indicates that ML and DL can improve cloud intrusion detection, although dataset imbalance, limited datasets, computational cost, false positives, and evolving attacks remain important challenges. The study argues that AI should complement rather than replace conventional security controls and human expertise. A resilient architecture must combine intelligent detection with zero trust, defense-in-depth, secure configuration, continuous monitoring, automated containment, backup, recovery, governance, and human oversight.