The Cluster spacecraft crossed the magnetopause at the duskward flank of the tail while the European Incoherent Scatter (EISCAT) radars and magnetometers observed the ionosphere during a sequence of intense substorm-like geomagnetic activity in October 2003. We attempt to estimate the local and global energy flow from the magnetosheath into the magnetotail and the ionosphere under these extreme conditions. We make for the first time direct observational estimates of the local solar wind power input using Cluster measurements. The global power input based on Cluster observations was found to be between 17 and 40 TW at the onset of the substorm intensification. However, spacecraft observations and global modelling of the magnetotail suggest that it is most probably closer to 17 TW. This is more than two times lower than the predicted $epsilon$ parameter value (37 TW). Energy deposition in the ionosphere has been estimated locally with EISCAT and globally with the assimilated mapping of ionospheric electrodynamics (AMIE) technique. The amount of the global solar wind power input (17 TW) that is dissipated via Joule heating in the ionosphere is found to be 30%. The corresponding ratio based on empirical estimates is only 3%. However, empirical proxies seem to underestimate the magnitude of the Joule heating rate as compared to AMIE estimates (~ a factor 4) and the $epsilon$ parameter is more than twice as large as the Cluster estimate. In summary, the observational estimates provide a good balance between the energy input to the magnetosphere and deposition in the ionosphere. Empirical proxies seem to suffer from overestimations ($epsilon$ parameter) and underestimations (Joule heating proxies) when pushed to the extreme circumstances during the early main phase of this storm period.