If the vendor's representative beam data (RBD) for TrueBeam linear accelerators are to be valid for use in clinical practice, the variations in the beam data used for beam modeling must be small. Although a few studies have reported the variation of the beam data of the TrueBeam machines, the numbers of machines analyzed in those studies were small. In this study, we investigated the variation in the beam data for 21 TrueBeam machines collected from 17 institutions with their agreement. In the exponential regions, the percent depth dose (PDD) values showed very small variation, <1% for all the photon energies analyzed. Similarly, the off‐center ratio (OCR) values also showed small variation for all energies. In the field regions, the standard deviations of the values of dose difference (DD) between the data for each machine and the study average were <1% for field sizes ≥100 × 100 mm2. The maximum distance‐to‐agreement from the average data was <0.5 mm in the penumbra regions. The output factor (OPF) values also showed very small variation (<1%) for all energies and field sizes. Both the PDD and OCR of the average study data showed good agreement with the vendor's RBD for field sizes ≥100 × 100 mm2. The OPF of the average study data also showed good agreement with the vendor's RBD for all field sizes. However, although all the institutions used ionization chambers with similar cavity volumes, the 30 × 30 mm2 field size showed large DD variations (≥2%) in OCR in the field regions. We conclude that the intermachine variability of TrueBeam linear accelerators was very small except for small field dosimetry, supporting the validity of the use of the RBD for clinical applications. The use of the vendor's RBD might greatly facilitate the quick installation of a new linear accelerator.