Large swings in the demand for content are commonplace within the Internet. When a traffic hotspot happens, however, there is a delay before measures such as heavy replication of content can be applied. This paper investigates the potential for predicting hotspots sufficiently far, albeit shortly, in advance, so that preventive action can be taken before the hotpot takes place. Performing accurate load predictions appears to be a daunting challenge at first glance, but this pape r shows that, when applied to web-server page-request traffic, even elementary prediction techniques can have a surprising forecasting power. We first argue this predictability from principles, and then confirm it by the analysis of empirical data, which reveals that large server overloads can often be seen well in advance. This allows steps to be taken to reduce substantially the degradation of service quality.