Earthquake early warning systems (EEWSs) are considered to be one of the most effective means for seismic risk mitigation, in terms of both losses and societal resilience, by releasing an alarm immediately after an earthquake occurs and before strong ground shaking arrives the target sites to be protected. To gain experience for the National System for Fast Seismic Intensity Report and Earthquake Early Warning project, we deployed a hybrid demonstration EEWS in the Sichuan–Yunnan border region with micro‐electro‐mechanical system‐based sensors and broadband seismographs and low‐latency data transmission. In this study, we described the structure of this EEWS and analyzed its performance in the first 2 yr from January 2017 to December 2018. During this test period, the EEWS detected and processed a total of 126 3.0+ earthquakes, with excellent epicentral location and magnitude estimation. The average location and magnitude estimation errors for the first alert were and , respectively. For the earthquakes that occurred inside and outside the hybrid network, the first alert was generated and after the origin time (OT), respectively. We analyzed the performance of the EEWS for the 31 October 2018 5.1 earthquake, because it was the largest event that occurred inside the hybrid network during the test period. The first alert was obtained at 7.5 s after the OT, with a magnitude error of 0.1 magnitude unit, a location error of about 1 km, and a depth error of 8 km. Finally, we discussed the main differences between the EEWS’s estimates and the catalogs obtained by the China Earthquake Network Center, and proposed improvements to reduce the reporting time. This study demonstrated that we constructed a reliable, effective hybrid EEWS for the test region, which can provide sufficient support for the design of the National EEWS project.