A probability density function of demand for automated external defibrillators (AED's), which is based on the distribution of demand points by kernel density estimation method, is estimated from the data on occurrences of cardiopulmonary arrests. The probability that someone can be saved by AED's is formulated using probability of survival to hospital discharge. "Supply effect" is expressed as demand density multiplied by this probability of saving, and the locations of AED's are optimized by maximizing supply effect in the entire region. The results show that optimal locations tend to cover the center of the city and that the supply effect is approximately proportional to the number of AED's.