Nvidia unveiled its latest AI supercomputer, the DGX GH200, at Computex in Taipei. The system is composed of 256 Grace-Hopper superchips connected via NVLink, and is tuned for memory-intensive AI models such as natural language processing, recommender systems, and graph neural networks. The DGX GH200 features 18,432 cores, 256 GPUs, and a claimed 144TB of “unified” memory, with 20TB of that being the super-speedy HBM3 used to store model parameters.