We propose and numerically investigate a novel direct route to produce multi-terawatt femtosecond selfcompressed 10 µm laser pulses suitable for the next generation relativistic laser-plasma studies including laser-wakefield acceleration at long wavelengths. The basic concept involves selecting an appropriate isotope of CO2 gas as a compression medium. This offers a dispersion/absorption landscape that is shifted in frequency relative to the driving CO2 laser used for 10 µm picosecond pulse generation. We show numerically that as a consequence of low losses and a broad anomalous dispersion window, a 3.5 ps duration pulse can be compressed to ~300 fs while carrying ~7 TW of peak power in less than 7 m. An interplay of self-phase modulation and anomalous dispersion leads to a ~3.5 times compression factor, followed by the onset of filamentation near the cell exit to get below 300 fs duration.