Accelerated life tests(ALT) are frequently used to obtain information on the lifetimes of reliable products quickly. Most existing researches on accelerated life testing are concerned with the case of a single stress. In this research, we consider ALT``s with two stresses which are expected to be able to obtain information on failure times more quickly than those with a single stress. Optimal ALT plans with two stresses are developed under Type I censoring when the lifetime is lognormally distributed. The relationship between the product life and stresses is assumed to follow generalized Eyring relationship. Optimal ALT plans with two stresses are compared with optimal ALT plans with a single stress. As a measure of comparison, we adopt the asymptotic variance of the estimate of the 10th percentile of the lifetime distribution at the use condition. Computational results are tabulized so that they may be readily available to the experimenters.