Gas migration velocity impacts the planning of pressurized mud cap drilling (PMCD) as it plays a pivotal role in the selection of fluid volumes and logistics. A pilot-scale experimental investigation of gas migration under downhole conditions (up to 3,600 psi, 240°F) in water, oils, and low-density drilling fluids is presented. While bubble-rise phenomena have been studied at near atmospheric pressures, the experimental setup and measurement method for high-temperature, high-pressure gas migration is rare. Experiments were performed using three test apparatuses: two separate pressurized lengths of 3-inch pipe, one 10-ft long and the other 18-ft long, as well as a unique high-pressure, high-temperature rotating test section (RTS). The RTS is 10-ft long, having a 6 inch × 4 inch eccentric annular geometry with the inner pipe capable of rotation. The inclination of all test sections can be varied. Gas was injected from the bottom through either a 1/8-inch diameter pressurized-injection port or a liquid-gas swap mechanism i.e. zero-velocity injection. Gas migration was recorded using a camera system or gamma-ray densitometers (GRDs). Some of the key results and insights from the testing are: (1) the gas migration rate and bubble length decrease with an increase in pressure, (2) the gas migration rate is higher in inclined vs. vertical sections, (3) bubble breakup occurs as pressure increases and interfacial tension decreases, (4) the inclination of the fluid column delays bubble breakup, and (5) high viscosity hinders bubble breakup. A key observation from the testing was that Taylor bubbles that may form during the initial phase of gas entering the annulus are likely to break up under downhole conditions of high pressure, low interfacial tension, and typical field mud viscosities, resulting in much lower gas migration rates during PMCD than the commonly used industry correlations. Another observation was that the practical length limitation of the test articles prevents us from observing the full evolution of gas bubble breakup. The results seen here are in line with our previous simulation work (Samdani et al., 2021, 2022).