A cloud virtual machine equipped with one or more Graphics Processing Units (GPUs) for accelerated computing workloads. Think of it like upgrading from a regular car to a race car — GPU instances provide massive parallel processing power needed for machine learning training, 3D rendering, video encoding, and scientific simulations. AWS offers P and G instance families, Azure provides NC and ND series, GCP has A2 and G2 machine types, and OCI offers GPU shapes with NVIDIA GPUs.
A machine learning team uses AWS p5.48xlarge GPU instances with 8 NVIDIA H100 GPUs to train a large language model, reducing training time from weeks to days compared to CPU-only instances.
All major cloud providers offer GPU-accelerated virtual machines optimized for ML training, inference, rendering, and HPC workloads.