The U.S. Food and Drug Administration (FDA) suggests the “Reference‐Scaled Average Bioequivalence” (RSABE) method in the average bioequivalence (ABE) study of highly variable drugs. The classic sample size estimation method for grouping sequential design (GSD) of the RSABE method is a single‐stage sample size multiplied by the inflation factor. This study proposed a new method for calculating the sample size for the GSD of the RSABE: using simulation experiments directly. In this work, our focal point is on a two‐stage GSD that adheres to the Pocock guideline, comprising a single interim analysis and a final analysis. We consider that the sample size of the two stages is equal; that is, the interim analysis is carried out at 50% of the information fraction. Extensive Monte Carlo simulations have shown that the new method is more accurate in estimating sample size than the inflation factor method, and the type I error rate is controlled below 5% in all conditions. In 90% power semireplicate studies, the average sample size required for the Pocock design is only 40% to 90% of the single‐stage design sample size.