Suction bucket foundations are an innovative, cost-effective, and environmentally sustainable solution for offshore wind turbine installations. Their deployment relies on generating suction pressure inside the bucket to achieve full embedment into the seabed. However, a critical challenge during installation is piping erosion, a process where particle fluidization beneath the bucket wall tip leads to a sudden drop in suction pressure, potentially causing installation failure. Despite its practical significance, the underlying physics governing piping erosion remain insufficiently understood due to the complex interplay between hydrodynamic forces and sediment transport. To bridge this knowledge gap, we employ a three-dimensional, fully-resolved coupled lattice Boltzmann method - discrete element method simulation to investigate the fundamental mechanisms driving piping erosion. The approach allows the capture of grain-scale interactions with high fidelity, enabling the identification of key parameters influencing erosion onset and progression. However, achieving physically representative problem sizes—comprising hundreds of thousands of grains—requires immense computational resources due to the high-resolution nature of the simulations. To address this computational challenge, simulations are performed on the LUMI supercomputer, leveraging hundreds of GPUs to execute large-scale, high-fidelity calculations. We present simulation results that showcase the method’s capability to resolve the intricate physics of piping erosion, offering new insights into the conditions that trigger erosion and its impact on suction bucket performance. These findings contribute to a deeper understanding of suction bucket installation dynamics and support the development of optimized, failure-resistant foundation designs for offshore wind applications.