A thermally-assisted recording (TAR) head structure has a semiconductor substrate as the head carrier with a vertical-cavity surface-emitting laser (VCSEL) on its front surface, a TAR head formed directly on the VCSEL, and a highly reflecting third mirror on its back surface. The semiconductor substrate serves as an extended cavity for the VCSEL. The TAR head is fabricated on the outer surface of the VCSEL in the same manner as proposed for fabrication of a TAR head on a conventional slider. The TAR head includes a conventional read head and write head, and an optical waveguide with a grating coupler and a near-field transducer (NFT). The laser radiation is output through a partially reflecting output mirror of the VCSEL through the front surface to the grating coupler, which turns the incoming laser radiation 90 degrees and directs it into the waveguide from where it is directed to the NFT.