The Compact Muon Solenoid (CMS) experiment at the Large Hadron Collider (LHC) studies proton-proton collisions at a centre-of-mass energy of 13 TeV. With the LHC colliding proton bunches every 25 nanoseconds, the volume and rate of raw data produced by the detector are much larger than what can be read out, recorded, and reconstructed. Therefore, an efficient trigger system is required to identify events of interest in real time and to reduce the rate of events to a manageable level for later software reconstruction. The CMS trigger system consists of two processing stages, a level-1 (L1) hardware trigger and a high level software trigger. The current L1 trigger decision relies solely on calorimetric and muon system information. During the High Luminosity LHC (HL-LHC) era, the instantaneous luminosity of the collider is expected to increase by approximately an order of magnitude, resulting in a significantly larger number of collisions per bunch crossing than observed in the current run. In order to preserve physics performance under such highly challenging conditions, the L1 trigger system must be upgraded to accommodate the use of silicon tracker data. A new CMS L1 Tracking Trigger will reconstruct tracks with transverse momentum above 2 GeV/c at the LHC bunch crossing rate and within a tight latency budget of approximately 4 $\mu$s. In this article we provide an overview of the three architectures currently studied by the CMS collaboration for the future L1 Track Trigger system.