Large-scale Direct Numerical Simulations (DNS) of turbulent combustion have become realizable with the advent of supercomputers. Given their immense computing capacity, supercomputers are playing a pivotal role in the development of next-generation clean gas turbine engine combustors powered by hydrogen combustion. This is being enabled through sophisticated numerical simulation methods like DNS, which can be used to elucidate and predict the complex flow and transport phenomena involved in the turbulent combustion of hydrogen inside gas turbine combustors. Therefore, a large-scale DNS of an actual turbulent lean-premixed swirling hydrogen flame is performed in this research, using an in-house code FK3, which employs a hybrid parallel architecture combining Message Passing Interface (MPI) and Open Multi-Processing (OpenMP). The DNS domain is discretized by a non-uniform Cartesian grid with approximately 6 Billion grid points (1800 × 1800 × 1800). The massively parallel DNS consumes 11.5 Million node-hours by parallel computation using 8000 nodes (with 25 CPUs per node, i.e., a total of 200,000 CPUs) on the supercomputer Fugaku at RIKEN, Japan. The generated DNS database (Big Data) provides an accurate spatiotemporal description of the flame’s thermochemical state and shows that the differential diffusion of light chemical species and heat, which influences the local flame structure and flame propagation speed, is well predicted. The Big Data is also interrogated for the presence of different localized combustion modes, such as premixed and non-premixed flame zones which are caused by reactant inhomogeneities arising from differential diffusion effects, mixing/entrainment in the shear layers and strain induced by the turbulent swirling motion of the flame.