In order to assess the moisture content of wood chips on an industrial scale, readily applicable techniques are required. Thus, near infrared (NIR) spectroscopy was used to estimate moisture in wood chips by means of partial least squares regressions. NIR spectra were obtained in spectrometer with an integrating sphere and optical fiber probe, on the longitudinal and transverse surface of Eucalyptus wood chips. The specimens had their masses and NIR spectra measured in 10 steps during drying from saturated to anhydrous condition. Principal Component Analysis was performed to explore the effect of moisture of wood chip on NIR signatures. The values of moisture content of chips were associated with the respective NIR spectra by Partial Least Squares Regression (PLS-R) and Partial Least Squares Discriminant Analysis (PLS-DA) to estimate the moisture content of wood chips and its moisture classes, respectively. Model developed from spectra recorded on the longitudinal face by the integrating sphere method presented statistics slightly better (R²cv = 0,96; RMSEcv = 7,15 %) than model based on optical fiber probe (R²cv = 0,90; RMSEcv = 11,86 %). This study suggests that for calibration of robust predictive model for estimating moisture content in chips the spectra should be recorded on the longitudinal surface of wood using the integrating sphere acquisition method.