The present disclosure provides a CMOS-MEMS device structure. The CMOS-MEMS device structure includes a sensing substrate and a CMOS substrate. The sensing substrate includes a bonding mesa structure. The CMOS substrate includes a top dielectric layer. The sensing substrate and the CMOS substrate are bonded through the bonding mesa structure, and the bonding mesa structure defines a bonding gap between the CMOS substrate and the sensing substrate.