Driven by the rapid evolution of integrated circuits toward higher power density, 2.5D/3D integration, chiplet-based architectures, and heterogeneous packaging, thermal management has emerged as a primary constraint on performance, reliability, and lifetime. This review provides a structured synthesis of the state of the art in chip thermal design by i) organizing numerical models across length scales, ii) summarizing experimental temperature and thermophysical-property characterization methods, and iii) critically analyzing the major bottlenecks that limit predictive accuracy under extreme heat-flux conditions. On the modeling side, we compare fast architecture-level approaches (equivalent RC thermal networks) with high-fidelity package- and system-level solvers (finite-element/finite-volume methods and conjugate heat-transfer CFD), and extend the discussion to micro-/nanoscale heat transport where non-Fourier effects become important, including phonon Boltzmann transport formulations and molecular dynamics for interfacial thermal boundary resistance. On the measurement side, we summarize the operating principles, spatiotemporal resolution, and applicability of infrared thermography, thermoreflectance microscopy, Raman thermometry, and embedded on-chip sensors, and highlight how these techniques are used to calibrate power maps, boundary conditions, and interface parameters for simulation–experiment closed-loop validation. Based on the literature, we identify recurring challenges: the prohibitive cost of full-chip transient multi-physics simulation, uncertainties in material properties and interfaces, limited simultaneous spatial and temporal resolution in experiments, and the approaching physical and practical limits of conventional air/heat-pipe cooling. Finally, we discuss emerging directions that can address these gaps, including AI-accelerated surrogate modeling and physics-informed learning for rapid thermal prediction, embedded microchannel and two-phase cooling for ultra-high heat flux, advanced high-thermal-conductivity interface/packaging materials, and multi-physics co-design that couples device, package, and cooling-system optimization.