Objective To estimate future excess mortality attributable to cold spells in Guangzhou, China. Methods We collected the mortality data and metrological data from 2009-2013 of Guangzhou to calculated the association between cold spell days and non-accidental mortality with GLM model. Then we projected future daily average temperatures (2020- 2039 (2020s),2050- 2069 (2050s),2080- 2099(2080s)) with 5 GCMs models and 2 RCPs (RCP4.5 and RCP8.5) to identify cold spell days. The baseline period was the 1980s (1980-1999). Finally, calculated the yearly cold spells related excess death of 1980s, 2020s, 2050s, and 2080s with average daily death count of non-cold spell days, exposure-response relationship, and yearly number of cold spell days. Results The average of daily non-accidental mortality in Guangzhou from 2009 to 2013 was 96, and the average of daily average was 22.0 ℃。 Cold spell days were associated with 3.3% (95% CI: 0.4%-6.2%) increase in non-accidental mortality。In 1980s, yearly cold spells related deaths were 34 (95%C/: 4-64). In 2020s, the number will increase by 0-10; in 2050s, the number will increase by 1-9; and in 2080s, will increase by 1-9 under the RCP4.5 scenario. In 2020s, the number will increase by 0-9; in 2050s, the number will increase by 1-6; and in 2080s, will increase by 0-11 under the RCP8.5 scenario. Conclusion The cold spells related non-accidental deaths in Guangzhou will increase in future under climate change.