Permafrost thaw subsidence, a key indicator of permafrost degradation, remains poorly quantified or understood. It is particularly challenging to detect and measure surface subsidence due to the loss of subsurface ice over a large area because it usually develops gradually, over several years or decades. Here we utilize the persistent scatterer interferometric synthetic aperture radar (PSI) approach to remotely measure gradual surface subsidence on Eboling Mountain in the northeastern region of the Qinghai‐Tibet Plateau, where thermal erosion gullies are well developed. Most of the previous multitemporal interferometric synthetic aperture radar studies on permafrost used the small baseline subset method. By contrast, the PSI approach benefits from the full spatial resolution and is less affected by temporal or geometric decorrelation. In the PSI analysis, we incorporate a piecewise elevation change model that includes periodic subsidence/uplift because of its seasonally varying components as well as its linear subsidence trends. Applying this permafrost‐designated PSI algorithm to 17‐L band ALOS‐1 PALSAR images taken between 2006 and 2011, we find that both the thermal erosion gullies and the surrounding regions (within about 300 m) subside gradually. The subsidence trends range from 0.3 to 3 cm/yr. This suggests that permafrost areas near the gullies are more vulnerable to gradual thawing and degradation. This study demonstrates the potential of using PSI to study permafrost thaw processes and of assessing its impacts over vast areas on the Qinghai‐Tibet Plateau and in the Arctic. Key Points Persistent scatterer interferometry is used to map gradual permafrost thaw subsidence with the full spatial resolution by exploiting all the repeat SAR acquisitions A piecewise elevation change model is introduced to capture the dynamic processes of the active layer and the underlying permafrost Ground surface subsidence, ranging from 0.3 to 3 cm/yr, occurred both inside the visible thermal erosion gullies and their surroundings