Massive star evolution strongly depends on their multiplicity properties at birth. Characterizing the multiplicity of young massive stars has seen some significant progress over the past years, thanks to the advent of optical/IR high-angular resolution facilities. Most main sequence massive stars ( 70%) belong to short-period binaries, a fact that does not reflect the binary parameters measured among populations of newly born massive stars. To bridge the gap between these two regimes, we need to obtain strong constraints on the multiplicity properties at birth and the origin of the pairing mechanism. Still, being deeply embedded in their natal envelope, a definitive observational sequence for their formation is lacking. A worthwhile compromise is to spatially resolve young binaries by means of near-IR high-angular resolution. Being one of the youngest clusters in our Galaxy, M17 is an unprecedented laboratory where pre-main sequence binaries can be caught. I will describe how the mas resolution scale achieved with VLTI/GRAVITY provides an unparalleled opportunity to investigate binarity among a population of young massive stellar systems. From the interferometric model fitting of visibility amplitudes and closure phases, I will present statistics on young massive stars multiplicity tracing 1-100mas in M17, including multiplicity and companion fraction. These results will be compared to other studies involving different star populations and covering different mass regimes. Finally, I will discuss the connection with the current star formation theories and relate with the migration scenario.