The three-year-lasting earthquake swarm in the Noto Peninsula, Japan, led to the 2024 Noto earthquake (moment magnitude 7.5). To reveal structural controls on the swarm evolution and the large earthquake generation, we imaged the three-dimensional S-wave velocity structure beneath the Noto Peninsula using ambient noise surface wave tomography with dense seismic observations. This dataset provides the cross-correlation functions of the densely observed ambient seismic noise and the estimated three-dimensional S-wave velocity structure model. We computed the multicomponent cross-correlation functions for the pairs of 22 seismic stations (composed of 12 seismic nodes and 10 short-period permanent stations). The observation data spanned 32 days, from October to November 2023. The cross-correlation functions in the vertical-vertical, vertical-radial, radial-radial, and transverse-transverse components yielded the estimates of dispersion curves of Rayleigh and Love waves. The dispersion curves of the fundamental-mode Rayleigh and Love waves in the frequency range of 0.10–0.45 Hz were used to estimate isotropic S-wave velocity structure via two-dimensional phase velocity tomography and a series of local 1D inversions. Our results show a high-velocity body collocated with the major slip zone of the 2024 Noto earthquake, a region the preceding swarm avoided. This observation strongly suggests that the structural heterogeneity beneath the Noto Peninsula controlled the swarm evolution and the generation of the 2024 Noto earthquake.