Physics-informed digital twins represent a decisive step toward understanding the true behavior of bridge structures throughout their entire lifecycle. This paper presents the field validation of a high-fidelity, physics-informed digital twin for a prestressed concrete highway bridge commissioned by ASFINAG Austria. The system has been in continuous operation for more than four years, covering all phases from construction and early-age behavior to long-term service conditions. The digital twin integrates distributed fiber-optic sensing, nonlinear finite element analysis, and inverse parameter identification into a unified framework that continuously calibrates itself to the real bridge response. Over 7 km of fiber-optic cables are embedded in the deck, webs, piers, and foundations, providing distributed strain and temperature data. These data are processed through the cloud-based WeStatiX SHM platform, which performs automated model updating using physics-informed neural networks (PINNs) and operational modal analysis (OMA). This approach ensures that the digital twin remains consistent with the governing physics of the structure while adapting to changing conditions. Four years of continuous monitoring have revealed a significant deviation between actual and design-predicted behavior, particularly regarding creep, shrinkage, and thermal gradients. Daily nonlinear simulations have shown that these effects dominate the long-term deformation and stress redistribution of the structure. Validation through live load testing confirmed that measured and simulated strains agree within 1%, demonstrating the robustness of the self-calibrating process. The results provide one of the first long-term validations of an autonomous, physics-informed digital twin for bridge infrastructure. The system transforms SHM from data collection into predictive, knowledge-based asset management, enabling objective condition assessment and proactive maintenance planning.