NVIDIA announced its new Ampere architecture with the A100 graphics processing unit (GPU). The A100 GPU is offered as part of an extremely high density DGX board and is optimized for artificial intelligence (AI)/machine learning (ML) workloads. This represents a major step forward in performance, but also an unprecedented challenge in terms of power density and cooling. It is also a big bet on an AI strategy defined by very large multipurpose models, centralized model training, and cloud services.