Nearly 50 years ago Bjerknes1 suggested that the character of large-scale air–sea interaction over the mid-latitude North Atlantic Ocean differs with timescales: the atmosphere was thought to drive directly most short-term—interannual—sea surface temperature (SST) variability, and the ocean to contribute significantly to long-term—multidecadal—SST and potentially atmospheric variability. Although the conjecture for short timescales is well accepted, understanding Atlantic multidecadal variability (AMV) of SST2, 3 remains a challenge as a result of limited ocean observations. AMV is nonetheless of major socio-economic importance because it is linked to important climate phenomena such as Atlantic hurricane activity and Sahel rainfall, and it hinders the detection of anthropogenic signals in the North Atlantic sector4, 5, 6. Direct evidence of the oceanic influence of AMV can only be provided by surface heat fluxes, the language of ocean–atmosphere communication. Here we provide observational evidence that in the mid-latitude North Atlantic and on timescales longer than 10 years, surface turbulent heat fluxes are indeed driven by the ocean and may force the atmosphere, whereas on shorter timescales the converse is true, thereby confirming the Bjerknes conjecture. This result, although strongest in boreal winter, is found in all seasons. Our findings suggest that the predictability of mid-latitude North Atlantic air–sea interaction could extend beyond the ocean to the climate of surrounding continents.