This study investigates the neural representation of internal speech in the supramarginal gyrus (SMG) of two participants with tetraplegia. The participants were implanted with microelectrode arrays in the SMG and primary somatosensory cortex (S1) to record neural activity during internal and vocalized speech tasks. The results show significant neural representation of both internal and vocalized speech at the single-neuron and population levels in the SMG. Decoding accuracies for internally spoken and vocalized words were 55% and 24% offline, and 79% and 23% online, respectively. The study also found evidence of shared neural representations between internal speech, word reading, and vocalized speech processes in one participant. Additionally, the SMG represented words and pseudowords, indicating phonetic encoding. The activity in S1 was modulated by vocalized speech but not internal speech, suggesting no articulatory movements during internal speech production. These findings provide a proof-of-concept for a high-performance internal speech brain-machine interface (BMI).This study investigates the neural representation of internal speech in the supramarginal gyrus (SMG) of two participants with tetraplegia. The participants were implanted with microelectrode arrays in the SMG and primary somatosensory cortex (S1) to record neural activity during internal and vocalized speech tasks. The results show significant neural representation of both internal and vocalized speech at the single-neuron and population levels in the SMG. Decoding accuracies for internally spoken and vocalized words were 55% and 24% offline, and 79% and 23% online, respectively. The study also found evidence of shared neural representations between internal speech, word reading, and vocalized speech processes in one participant. Additionally, the SMG represented words and pseudowords, indicating phonetic encoding. The activity in S1 was modulated by vocalized speech but not internal speech, suggesting no articulatory movements during internal speech production. These findings provide a proof-of-concept for a high-performance internal speech brain-machine interface (BMI).