Velda Redefines Serverless GPU Computing for AI Developers and Batch Workloads

Velda is redefining serverless GPU computing for AI developers with a platform that lets teams launch training jobs, batch inference pipelines, and distributed workloads directly from their local development environment using a single command prefix. In San Francisco on June 11, 2026, Velda introduced the vrun command, which prepends to any script to offload heavy compute to cloud GPUs, including H200 instances, without changing a line of code. There are no image builds, no manifest files, and no infrastructure configuration required, preserving the local development experience while scaling compute in the cloud. This approach eliminates the traditional overhead of packaging environments, defining clusters, and provisioning resources, addressing a core bottleneck in modern AI development. This platform supports robotics and physical AI workloads, enabling simulations and embodied AI alongside language model training and batch inference. Velda operates in two tiers: Velda Cloud, a managed offering that provides browser-based VS Code, GPU compute, and a free monthly credit; and an Enterprise tier with self-hosted or dedicated deployment options plus onboarding sessions. The emphasis on developer experience and frictionless scale differentiates Velda as compute demand for GPUs continues to outpace tooling. The company did not disclose pricing specifics beyond the free credit, but it positions Cloud for individuals and small teams and Enterprise for larger organizations with stricter security needs.