Under global warming, increased wildfires activities now pose unprecedented risks to ecosystems and human societies. The key question is no longer if wildfire activity will increase, but how—whether through gradual intensification or a threshold shift into qualitatively different regimes with catastrophic events. Here we introduce an AI framework for more observationally constrained wildfire risk assessment in the future, revealing that socioeconomic development policies are the primary determinant shaping the divergent wildfire regime shift. Under the very high-emission scenario (SSP585), global wildfires will not only increase but also face a qualitative transformation characterized by peak intensification and extension of wildfire seasons. While land use change has driven recent global wildfire declines, climate change will drive the future wildfire activity. Vapor pressure deficit (VPD) emerges as the overwhelmingly dominant driver, responsible for both overall increases and seasonal peak intensification. Although land use change has relatively small direct impact, it plays a crucial role in providing the background fuel availability that amplifies VPD’s drying effects, especially in Mediterranean, Amazon, Siberia and North America. This interaction challenges conventional focus on climate indicators alone and provides a mechanistic understanding for identifying high-risk regions and developing targeted adaptation strategies.

widefire,AI