By using 64,270 daily observations from a large hospital in Guangzhou between 2017 and 2019, we analyzed the impact of extreme temperature on the health of newborns via OLS regression with time fixed effect. Given that the short-term temperature change can be regarded as exogenous and random, solving the potential endogenous problem is critical. We find that extreme temperature negatively affects the health of newborns. The Apgar score, an index for evaluating neonatal health, decreases by 0.008 (0.029%) when the duration of extreme temperature events increases by a day. A series of robustness checks verify the reliability of this negative effect. Extreme temperature also has a particularly serious effect on the health of newborns whose mothers have poor education. By gradually extending the observation period, we find that the effect of extreme temperature on neonatal health is mainly concentrated 1-6 weeks before delivery, whereas the effect of extreme temperature on hospitalization cost is mainly concentrated 4-8 weeks before delivery. This paper provides a valuable reference for evaluating the health and social costs of extreme weather, and our findings are conducive to the construction of climate resilient health systems, especially in Guangzhou.