Nvidia unveiled the DGX H300 platform to accelerate training and inference for extremely large AI models, offering higher efficiency and faster interconnect performance for data centres. The system pairs advanced H-series GPUs with software optimisations designed to shorten model training cycles and reduce overall TCO for hyperscalers and enterprise AI teams. Nvidia said the platform targets workloads such as large language models, multimodal systems and drug discovery research. Analysts noted the launch intensifies competition in AI hardware and will be closely watched for adoption by cloud providers and major research labs.