A new method for recording above 100 Gb/in.2 is discussed. We call this method “hybrid recording,” a form of thermally-assisted recording that combines thermo-magnetic writing and magnetic reading. In order to increase the stability of the recorded information, writing is carried out at an elevated temperature on a medium with a very high coercivity at room temperature. In our proposal write and read heads with extremely narrow trackwidths are used, so the trackwidth is not determined by the optical spot size and the written bits have a rectangular shape, in contrast to the schemes proposed by others. Preliminary experiments are shown. The applicability of today’s granular and MO type media for hybrid recording is discussed. It is calculated that hybrid recording on optimized media can give an increase of the areal density of a factor 2.9 in areal density or 7 dB (2.2×) medium SNR improvement in case of Poisson noise and 11 dB (3.4×) in case of transition noise. Practically a factor of about 2 in density is more realistic, pushing limiting densities for longitudinal recording to 100–200 Gb/in.2. Typical limitations at very high densities arise from heat dissipation in the head and thermal instability of the medium. Based on simplified model calculations including realistic limitations on medium, head and leads, and today’s practical limitations on electronics, comparisons are made between read heads containing a tunnel junction magnetoresistive (TMR) element and containing a giant magnetoresistive element with sense current in the plane (CIP-GMR) or perpendicular to the plane (CPP-GMR) of the sensor films. They show that the signal-to-noise ratio of TMR sensors for areal densities above 15 Gb/in.2 is - not advantageous over GMR sensors with sense current in the plane as long as the junction’s tunnel resistance is not drastically reduced to below 10 Ω μm2. The CPP-GMR heads are disadvantageous with respect to CIP-GMR heads until the highest densities, 300 Gb/in.2, considered. © 2000 American Institute of Physics.