For much of the last 66 million years, Earth's climate and carbon cycle have varied in synchrony on eccentricity timescales. The signals of global temperature and ice volume (δ¹⁸O) and the ocean's carbon budget (δ¹³C), recorded in the shells of ancient deep-sea organisms, rose and fell in phase for tens of millions of years. Yet the mechanism behind this persistent relationship has remained one of paleoceanography's enduring puzzles.

Our study addresses this mystery through new ultra-high-resolution data from IODP Site U1505 in the South China Sea, spanning the Miocene epoch (24–13 Ma), a critical period when only Antarctica was glaciated and the climate was warmer than today. Using benthic foraminiferal B/Ca and Cd/Ca ratios, we reconstructed deep-sea carbonate ion saturation (Δ[CO₃²⁻]) and phosphate concentration ([PO₄³⁻]) across this interval. Our results reveal that these proxies pulse with the same 405-kyr and 100-kyr eccentricity cycles that dominate the isotopic records. Critically, Δ[CO₃²⁻] varies in phase with δ¹⁸O and δ¹³C: at eccentricity maxima, when climate was warmest, deep waters became more corrosive and recorded lighter carbon isotopic values. This pattern is mirrored in global deep-sea CaCO₃ records, which show intensified dissolution at the same warm peaks.

To explain these observations, we developed a biogeochemical model of the global carbon cycle forced by orbital variations. The model, combined with our proxy data, reveals a coherent mechanism: eccentricity maxima amplify monsoon rainfall and intensify continental chemical weathering. This process delivers large fluxes of dissolved inorganic carbon (DIC) and alkalinity to the ocean. Coupled with higher sea levels, this influx drives massive carbonate precipitation and burial on the newly submerged continental shelves. This shallow-water carbonate sink draws alkalinity from the ocean, leaving deep waters depleted in carbonate ions and with a lighter δ¹³C signature.

Our model confirms that riverine input of isotopically light carbon, combined with the preferential burial of ¹³C-enriched carbonates on shelves, fully accounts for the ~1‰ fluctuations observed in deep-sea δ¹³C. This "weathering and carbonate burial" hypothesis provides a unified framework explaining the in-phase correlation between temperature, the carbon cycle, and deep-sea carbonate chemistry on eccentricity timescales.

Importantly, this mechanism was not confined to the Miocene. Similar eccentricity-paced covariation in benthic isotopes and carbonate records from the Oligocene (~34–23 Ma) indicates that the same process was operative throughout unipolar-glaciated intervals. The system only fundamentally reorganized after ~6 Ma, when the onset of bipolar glaciation shifted the dominant carbon-cycle feedbacks from low-latitude weathering to high-latitude ice-sheet interactions, reversing the phase relationship between δ¹⁸O and δ¹³C.