Significance: Spectroscopic analysis of optical coherence tomography (OCT) data can yield added information about the sample's chemical composition, along with high-resolution images. Typical commercial OCT systems operate at wavelengths that may not be optimal for identifying lipid-containing samples based on absorption features. Aim: The main aim of this study was to develop a 1200 nm spectroscopic OCT (SOCT) for the classification of lipid-based and water-based samples by extracting the lipid absorption peak at 1210 nm from the OCT data. Approach: We developed a 1200 nm OCT system and implemented a signal processing algorithm that simultaneously retrieves spectroscopic and structural information from the sample. In this study, we validated the performance of our OCT system by imaging weakly scattering phantoms with and without lipid absorption features. An orthogonal projections to latent structures-discriminant analysis (OPLS-DA) model was developed and applied to classify weakly scattering samples based on their absorption features. Results: The OCT system achieved an axial resolution of 7.2 m and a sensitivity of 95 dB. The calibrated OPLS-DA model on weakly scattering samples with lipid and water-based absorption features correctly classified 19/20 validation samples. Conclusions: The 1200 nm SOCT system can discriminate the lipid-containing weakly scattering samples from water-based weakly scattering samples with good predictive ability.