Previous estimates of the strength and the uncertainty of the observed Antarctic temperature trends assumed that the natural annual temperature fluctuations can be represented by an auto-regressive process of first order [AR(1)]. Here we find that this hypothesis is inadequate. We consider the longest observational temperature records in Antarctica and show that their variability is better represented by a long-term persistent process that has a propensity of large and enduring natural excursions from the mean. As a consequence, the statistical significance of the recent (presumably anthropogenic) Antarctic warming trend is lower than hitherto reported, while the uncertainty about its magnitude is enhanced. Indeed, all records except for one (Faraday/Vernadsky) fail to show a significant trend. When increasing the signal-to-noise ratio by considering appropriate averages of the local temperature series, we find that the warming trend is still not significant in East Antarctica and the Antarctic Peninsula. In West Antarctica, however, the significance of the trend is above \(97.4 \,\%\) , and its magnitude is between 0.08 and 0.96 °C per decade. We argue that the persistent temperature fluctuations not only have a larger impact on regional warming uncertainties than previously thought but also may provide a potential mechanism for understanding the transient weakening (“hiatus”) of the regional and global temperature trends.