Stress-driven reliability-by-design for heterogeneous integration: from packaging architectures to intelligent management

As traditional computing architectures encounter fundamental limits in performance and energy efficiency, heterogeneous integration has become crucial for ongoing progress, especially in emerging fields such as artificial intelligence, quantum computing, and photonic computing. The dense cointegration of components with different physical properties and manufacturing processes, however, introduces new reliability challenges based on multisource, cross-scale, and strongly coupled multiphysics phenomena. These issues pose a significant obstacle to fully realizing the potential of heterogeneous integration and reaching new computing frontiers. This review provides a systematic approach to understanding and addressing these reliability problems. We begin by analyzing the physics and interfacial challenges of integrating novel heterogeneous materials and devices to identify the microscopic causes and failure mechanisms. Building on this physical foundation, we then review reliability-focused innovations in packaging and system design, demonstrating how advanced interconnects and system-level strategies can support the scalability and energy efficiency of future computing systems. Finally, the review explores the emerging field of intelligent reliability, explaining how methods such as cosimulation, integrated sensing, and artificial intelligence-driven prediction and adaptive management can incorporate reliability as a key design element. This work outlines a strategic path toward developing future high-efficiency, high-throughput, and adaptable heterogeneous computing platforms.