Past coral range expansions suggest that high-latitude environments may serve as refugia, potentially buffering tropical biodiversity loss due to climate change. We explore this possibility for corals globally, using a dynamical metacommunity model incorporating temperature, light intensity, pH, and four distinct, interacting coral assemblages. This model reasonably reproduces the observed distribution and recent decline of corals across the Indo-Pacific and Caribbean. Our simulations suggest that there is a mismatch between the timescales of coral reef decline and range expansion under future predicted climate change. Whereas the most severe declines in coral cover will likely occur within 60–80 years, significant tropical coral range expansion requires centuries. The absence of large-scale coral refugia in the face of rapid anthropogenic climate change emphasises the urgent need to reduce greenhouse gas emissions, and mitigate non-thermal stressors for corals, both in the tropics and high-latitudes.