Currently, the satellite Microwave Sounding Unit (MSU/AMSU) datasets developed from three organizations—Remote Sensing Systems (RSS), the University of Alabama at Huntsville (UAH), and the NOAA Center for Satellite Applications and Research (STAR)—are often used to monitor the global long-term trends of temperatures in the lower troposphere (TLT), midtroposphere (TMT), total troposphere (TTT), troposphere and stratosphere (TTS), and lower stratosphere (TLS). However, the trend in these temperatures over China has not been quantitatively assessed. In this study, the decadal variability and long-term trend of upper-air temperature during 1979–2018 from three MSU datasets are first evaluated over China and compared with the proxy MSU dataset simulated from homogenized surface and radiosonde profiles (EQU) at 113 stations in China. The regional mean MSU trends over China during 1979–2018 are 0.22–0.27 (TLT), 0.15–0.22 (TMT), 0.20–0.27 (TTT), 0.02–0.14 (TTS), and from −0.33 to −0.36 (TLS) K decade−1, whereas the EQU trends are 0.31 (TLT), 0.19 (TMT), 0.24 (TTT), 0.07 (TTS), and −0.26 (TLS) K decade−1. The trends from RSS generally show a better agreement with those from EQU. The trends from both MSU and EQU exhibit seasonal and regional difference with a larger warming in TLT in February and March, and stronger cooling in TLS from late winter to spring. The TLT and TMT over the Tibetan Plateau and northwestern China show larger warming trends. The variability from MSU and EQU agree well except TLT in Tibet and southern China. The major difference in regional mean temperatures over China between MSU and EQU is related primarily to the satellite instrument changes during 1979–98 and the radiosonde system changes in China in the 2000s.