The Cretaceous to early Paleogene (ca. 140–50 Ma) was characterized by a greenhouse baseline climate, driven by elevated concentrations of atmospheric CO2. Hypotheses for the elevated CO2 concentrations invoke an increase in volcanic CO2 production due to higher oceanic crust production rates, higher frequency of large igneous provinces, or increases in pelagic carbonate deposition, the last leading to enhanced carbonate subduction into the mantle source regions of arc volcanoes. However, these are not the only volcanic sources of CO2 during this time interval. We show here that ocean-continent subduction zones, manifested as a global chain of continental arc volcanoes, were as much as 200% longer in the Cretaceous and early Paleogene than in the late Paleogene to present, when a cooler climate prevailed. In particular, many of these continental arcs, unlike island arcs, intersected ancient continental platform carbonates stored on the continental upper plate. We show that the greater length of Cretaceous–Paleogene continental arcs, specifically carbonate-intersecting arcs, could have increased global production of CO2 by at least 3.7–5.5 times that of the present day. This magmatically driven crustal decarbonation flux of CO2 through continental arcs exceeds that delivered by Cretaceous oceanic crust production, and was sufficient to drive Cretaceous–Paleogene greenhouse conditions. Thus, carbonate-intersecting continental arc volcanoes likely played an important role in driving greenhouse conditions in the Cretaceous–Paleogene. If so, the waning of North American and Eurasian continental arcs in the Late Cretaceous to early Paleogene, followed by a fundamental shift in western Pacific subduction zones ca. 52 Ma to an island arc–dominated regime, would have been manifested as a decline in global volcanic CO2 production, prompting a return to an icehouse baseline in the Neogene. Our analysis leads us to speculate that long-term (>50 m.y.) greenhouse-icehouse oscillations may be linked to fluctuations between continental- and island arc–dominated states. These tectonic fluctuations may result from large-scale changes in the nature of subduction zones, changes we speculate may be tied to the assembly and dispersal of continents. Specifically, dispersal of continents may drive the leading edge of continents to override subduction zones, resulting in continental arc volcanism, whereas assembly of continents or closing of large ocean basins may be manifested as large-scale slab rollback, resulting in the development of intraoceanic volcanic arcs. We suggest that greenhouse-icehouse oscillations are a natural consequence of plate tectonics operating in the presence of continental masses, serving as a large capacitor of carbonates that can be episodically purged during global flare-ups in continental arcs. Importantly, if the global crustal carbonate reservoir has grown with time, as might be expected because platform carbonates on continents do not generally subduct, the greenhouse-driving potential of continental arcs would have been small during the Archean, but would have increased in the Neoproterozoic and Phanerozoic after a significant reservoir of crustal carbonates had formed in response to the evolution of life and the growth of continents.