Recent fire events in Alaskan tundra ecosystems have been identified as harbingers of climate change and have caused reassessment of more traditional thinking about fire activity in this high-latitude biome. Although some work has demonstrated the novelty of these fires and linked weather, climate, and environmental variables to their occurrence, our understanding of their long-term (multi-decadal and centennial-scale) and spatially broad-scale climatic controls remains preliminary. We address this gap by comparing paleofire records with relevant climate, vegetation, and environmental datasets to identify the most important controls of tundra fire and assess their relative importance (using machine learning techniques). Additionally, we assess likely causal relationships of these environmental variables and fire activity using structural equation models (SEMs). We find that atmospheric CO2 is the primary control of tundra fire, followed by summer temperature, and precipitation. We infer that atmospheric CO2 directly increases lightning ignition frequencies while also promoting warm, dry climatic conditions conducive to fire. This inference is supported by analyses of observational lightning and fire data and is further supported by recent work identifying lightning as the dominant control of tundra fire occurrence. These findings are alarming in light of rising atmospheric CO2 and climate projections.