The tabular model is an in-memory engine meaning that the data in the model (all models on the tabular server) is compressed and stored in the RAM while the server is running. The tabular server also stores data on the disk (storage for when the SSAS engine is not running)—however, the rule is generally that all queries against the model are performed on cached data. The exception to this, of course, is when there is not enough RAM for the engine to use, and paging occurs.
This is a stark contrast to Multidimensional Online Analytical Processing (MOLAP) storage, which stores data on the disk (notwithstanding aggregations that may have been cached). This storage requires an effective partitioning strategy as a performance consideration. That is, a strategy to physically store the data on the disk.