The ERA5 reanalysis dataset is used to characterize the principal patterns of annual and seasonal near-surface wind speeds in the Arctic from 1979 to 2023 by the empirical orthogonal function (EOF) method, together with their relationship with large-scale atmospheric circulation. The leading EOF (EOF1) accounts for about 14%–22% of annual and seasonal wind speed variability, and is composed of widely positive anomalies across most of the Arctic Ocean largely influenced by the positive Arctic Oscillation. Seasonally, the positive anomalies shift from the central Arctic Ocean in summer toward the Greenland and Norwegian Seas in winter, driven by the North Atlantic Oscillation (NAO). The second EOF (EOF2) explains 9.5% of annual wind speed changes and consists of a meridional dipole structure, i.e. positive anomalies over the Russian Arctic seas and negative anomalies over the Greenland and Norway Seas, mainly driven by the Arctic dipole and NAO. The time series of EOF2 shows a decadal switch from negative to positive anomalies at the beginning of the 21st century, related to the phase transition of the Pacific Decadal Oscillation and the Atlantic Multidecadal Oscillation.