DigitalOcean has rolled out basic observability metrics for GPU Droplets and DOKS clusters, making it easier to monitor and optimize AI workloads. Key performance and stability metrics are crucial when handling large-scale training and data processing. The new features provide real-time insights on NVIDIA and AMD GPUs, showing metrics like utilization, temperature, power consumption, and more, all accessible through the DigitalOcean Insights UI without any setup. The observability metrics are organized into five categories: Utilization helps you gauge GPU core and memory usage; Temperature monitors thermal conditions to prevent overheating; Power tracks consumption for performance analysis; Throttle identifies performance limitations caused by thermal or voltage issues; and Interconnect evaluates network interface performance. These metrics are enabled by default for new GPU Droplets and come at no additional cost with AI/ML Ready images. GPU Droplets are priced starting at $0.76 per GPU per hour, with flexible configurations available to suit different needs. The platform offers seamless integration with existing DigitalOcean projects, including Kubernetes, and comes with enterprise-grade SLAs and compliance standards. Future updates are planned to enhance the observability suite further, indicating a commitment to improving the GPU experience on the platform.