In this work, we are exploring methods to cluster cells in spreadsheets based on their similarity. This work relates to following two research topics in spreadsheets: layout inference and table identification. In literature, already exists a body of publications that discuss solutions for the aforementioned topics. However, they work on the level of columns, rows, and/or cells. Contrary, we opt for a solution that utilizes clusters of cells. Unlike columns and rows, clusters contain cells having most of the characteristics (features) in common. Intuitively, clusters can be treated as one solid fragment of the sheet, with a degree of certainty that all the cells in them have the same layout role. Thus, this technique could effectively increase the accuracy for the layout inference task. Additionally, it could result in faster execution. Unlike the exhaustive cell-by-cell approach, cell clustering minimizes the number of objects in the process