The next major frontier in the study of extrasolar planets is direct imaging detection of the planets themselves. With high-order adaptive optics, careful system design, and advanced coronagraphy, it is possible for an AO system on a 8-m class telescope to achieve contrast levels of 10-7 to 10-8, sufficient to detect warm self-luminous Jovian planets in the solar neighborhood. Such direct detection is sensitive to planets inaccessible to current radial-velocity surveys and allows spectral characterization of the planets, shedding light on planet formation and the structure of other solar systems. We have begun the construction of such a system for the Gemini Observatory. Dubbed the Gemini Planet Imager (GPI), this instrument should be deployed in 2010 on the Gemini South telescope. It combines a 2000-actuator MEMS-based AO system, an apodized-pupil Lyot coronagraph, a precision infrared interferometer for real-time wavefront calibration at the nanometer level, and a infrared integral field spectrograph for detection and characterization of the target planets. GPI will be able to achieve Strehl ratios > 0.9 at 1.65 microns and to observe a broad sample of science targets with I band magnitudes less than 8. In addition to planet detection, GPI will also be capable of polarimetric imaging of circumstellar dust disks, studies of evolved stars, and high-Strehl imaging spectroscopy of bright targets. We present here an overview of the GPI instrument design, an error budget highlighting key technological challenges, and models of the system performance.