We describe a GPU-enabled approach for real-time optical frequency comb spectroscopy in which data is recorded, Fourier transformed, normalized, and fit at data rates up to 2.2 GB/s. As an initial demonstration we have applied this approach to rapidly interrogate the motion of an optomechanical accelerometer through the use of an electro-optic frequency comb. However, we note that this approach is readily amenable to both self-heterodyne and dual comb spectrometers for molecular spectroscopy as well as photonic readout where the approach's agility, speed, and simplicity are expected to enable future improvements and applications.