NVIDIA open-sources Snapshot to cut GPU inference cold starts to seconds

On November 10, 2026, NVIDIA Developer Blog announced that NVIDIA Snapshot has been open-sourced. According to the post, Snapshot is a new open-source component of the NVIDIA DSX AI Factory Platform and is a checkpoint-and-restore system for GPU workloads on Kubernetes. It captures a fully initialized workload, including its CPU and GPU memory, and restores that state on compatible infrastructure, making initialized GPU workloads reusable across inference replicas so that startup and recovery run faster. The post explains that when inference demand spikes or a replica fails, available GPUs do not translate immediately into serving capacity: initializing an LLM workload, a process known as a cold start, can take minutes, which delays both scaling and recovery. By capturing and restoring already-initialized state, Snapshot is said to bring inference capacity online in seconds rather than minutes. The material does not cover the project's license, repository location, general availability date, or any follow-up developments.