DigitalOcean GPU-accelerated virtual machines equipped with NVIDIA H100 or A100 GPUs, designed for AI/ML training, inference, and other GPU-intensive workloads. GPU Droplets provide on-demand access to enterprise-grade GPUs without the capital expense of owning hardware.
A machine learning team uses GPU Droplets with NVIDIA H100 80 GB GPUs to fine-tune a large language model on their proprietary dataset. They spin up a cluster of GPU Droplets for a 48-hour training run, then terminate them — paying only for the compute time used.