Nvidia unveiled the DGX H300, a next-generation AI supercomputing platform designed to train trillion-parameter models with improved energy efficiency and interconnect performance. The system pairs advanced H-series GPUs with optimized software stacks to accelerate large-language-model training and inference for enterprise and research customers. Nvidia positioned the platform as a way to shorten model training cycles and reduce total cost of ownership for hyperscalers and large enterprises. Analysts say the DGX H300 will intensify competition in the high-performance AI hardware market and pressure OEMs to upgrade data-centre infrastructure.