Earthquakes and Tsunamis remain among nature’s most unpredictable and destructive phenomena, originating on geological fault systems close to densely populated regions and critical infrastructure. They may cause many casualties and hundreds of billions to trillions of dollars in losses even in well-prepared societies and cannot be deterministically predicted. The underlying mechanics is governed by highly nonlinear dynamics and vary in size over an extreme range of magnitudes. Frictional weakening and dynamic rupture processes may occur over meters and fractions of seconds, while fault-system loading and stress evolution unfold over centuries and across hundreds of kilometers. Tsunami generation adds complexity because it depends on the spatiotemporal pattern of seafloor deformation, shallow fault and sediment properties, Ocean-Solid Earth coupling and bathymetry. Despite dense seismic, geodetic, and remote-sensing observations and a growing catalog of well-recorded events, a predictive, physics-based understanding of earthquake rupture, ground shaking, and tsunami generation remains incomplete. This talk argues that we may enter a transformative era in earthquake and tsunami science, enabled by the convergence of high-performance computing, data-science, dense observational data streams, and large-scale physics-based forward and inverse modeling. I will outline how ensembles of 3D dynamic rupture, seismic wave propagation, and tsunami simulations can be combined with data assimilation and Bayesian inference to estimate not only “best-fit” scenarios, but also uncertainties and parameter trade-offs that matter for decision-making [e.g., 1-3]. I will present recent advances in physics-based earthquake and tsunami models, closely integrated with interdisciplinary observations and utilizing supercomputing. The output of such high-fidelity models can be used to construct time-dependent reduced-order models, a low-dimensional and accurate approximation of the expensive forward simulations, that is fast enough to rapidly evaluate new earthquake scenarios for early warning, rapid response and physics-based probabilistic hazard assessment.