This study investigates the potential of physics-informed neural networks (PINNs) as a surrogate modeling method for chromatographic systems. Our aim is to enhance computationally intensive applications, including process optimization, uncertainty quantification, and model predictive process control. PINNs provide a gray-box modeling approach that leverages automatic differentiation during the learning process to calculate solutions to partial differential equations (PDEs) without relying on classical numerical methods. Established by Raissi et al. [1], this method allows physical principles to be incorporated into the network as an additional part of the loss function. While traditional mechanistic models dominate chromatography simulations, hybrid models like PINNs are gaining increasing attention in the academic community due to their potential speed advantages over mechanistic approaches and greater data efficiency compared to other data-driven methods. However, it is anticipated that they may require longer training times and are difficult to set up due to the complex error function. To investigate these claims, we conducted an in silico study by developing a PINN using simulated chromatography data derived from a fully parameterized mechanistic model; this model's PDEs were discretized and integrated using the open-source tool CADET [2]. The results of our study highlight both the strengths and limitations of PINNs in this context.