AI Cloud
From rapid prototyping to foundation training to scalable inference
Compute
GB300 NVL72
New1x tray to 2+ racks · NVLink v5
Create, store, and distribute container images co-located with the compute that runs them
Minimize the data transfer overhead with persistent volumes. Attach, detach, and clone across GPU instances.
High I/O throughput and low latency for AI training and inference.
Only pay for what you use. Delete instantly. No egress fees to Verda compute.
Pay only for what you store. Co-located with compute across all regions.
| Storage type | GB per month | TB per month |
|---|---|---|
| Container registry | $0.2000 | $200.00 |
The registry is co-located with Verda compute resources. Image pulls complete in seconds, not minutes — and never touch the public internet.
Reference an image to launch Serverless containers, Batch jobs, or Inference endpoints on Verda in one click — shortening the path from build to production.
Efficient image delivery enables faster startup times across the entire ML model lifecycle, helping to launch quickly without the orchestration overhead.
Store container images securely with private repositories, controlled access, and encrypted data transmission. Credentials can be disabled or set to a limited duration.
Use version tags and immutable images to ensure consistent environments across development, training, and production.
Standard Docker Registry v2 API. Works with any OCI-compliant client — Docker, Podman, Crane, Buildah, Kaniko, Skopeo.
Three things. First, the registry lives in the same data center as Verda's GPUs, so pulls for multi-GB AI images are fast and don't cross the public internet. Second, deploys to Serverless Containers, Batch Jobs, and inference endpoints are one step — no mirroring, no third-party credentials inside your jobs. Third, public registries like Docker Hub throttle pulls; a private registry tied to your compute doesn't.
Repositories are private by default. You grant push, pull, or admin rights per repository to users or service accounts. Service-account credentials can be disabled or time-limited so CI pipelines, training runs, and external collaborators get exactly the access they need, for only as long as they need it.
Yes. Standard Docker Registry v2 API. Works with docker, podman, crane, Buildah, Kaniko, Skopeo, and any OCI client.
Built in Europe, trusted globally
From rapid prototyping to foundation training and scalable inference — on a single full-stack AI cloud