We test the effectiveness of photometric redshifts based upon galaxy spectral template fitting for X-ray luminous objects, using a sample of 65 sources detected by Chandra in the field of the Caltech Faint Galaxy Redshift Survey (CFGRS). We find that sources with quasar-dominated spectra (for which galaxy spectral templates are not appropriate) are easily identified, and that photometric redshifts are robust for the rest of the sources in our sample. Specifically, for the 59 sources that are not quasar-dominated at optical wavelengths, we find that the photometric redshift estimates have scatter comparable to the field galaxy population in this region. There is no evidence for a trend of increasing dispersion with X-ray luminosity over the range L X = 10 39 − 5 × 10 43 erg s −1 , nor is there a trend with the ratio of X-ray to optical flux, f X / f R . The practical implication of this work is that photometric redshifts should be robust for the majority (∼90%) of the X-ray sources down to f X ≈ 10 −16 erg s −1 cm −2 that have optical counterparts brighter than R ≈ 24. Furthermore, the same photometry can be easily used to identify the sources for which the photometric redshifts are likely to fail. Photometric redshift estimation can thus be utilized as an efficient tool in analyzing the statistical properties of upcoming large Chandra and XMM-Newton data sets and identifying interesting subsamples for further study.