
Your Ray platform just changed hands. Your roadmap should not.
Anyscale’s shift to Nscale leaves teams facing new vendor lock-in and the same missing governance controls. Domino delivers enterprise policy enforcement and multi-cloud flexibility, keeping your Ray workloads independent of any single cloud or GPU provider. Keep your compute open and your platform secure.
What the acquisition changes
Ray stays open source, but Anyscale's roadmap now serves Nscale's hardware revenue. Because the platform lacks native experiment tracking, model registries, and approval workflows, regulated teams are still stuck building, maintaining, and auditing that infrastructure manually.
How the architectures differ:
Anyscale, post-Nscale
Domino
Control plane you host yourself
None documented. Kubernetes deployment requires internet egress from Anyscale pods; on Azure the control plane is hosted in Azure.
Self-managed VPC on EKS, AKS, or GKE, on-premises and bare metal, or hybrid data planes.
Disconnected and government deployed
No documented configuration.
Yes. Running in DoD IL5 on AWS GovCloud today. Deployed into Iron Bank.
Why move Ray work to Domino?
Domino pairs ephemeral Ray clusters with native Kubernetes scaling and enterprise governance.
- Tiered Compute: Head and worker nodes run on separate, on-demand CPU/GPU tiers.
- Native Scaling & Serving: Uses Kubernetes HPA for scaling and Domino Endpoints for production inference.
- Persistent Data: Read-only, versioned Domino Datasets prevent teardown data loss and guarantee exact reproducibility.
- Unified Governance: Automatically audits every run across Ray, Spark, Dask, Open MPI, and Slurm.
A large U.S. mortgage finance institution runs Spark and Ray on Domino to standardize fee data across more than 174 million records, cutting one analysis from two hours to about a minute and test cycles by 90 percent.
Governed Ray execution on any infrastructure
Get built-in compliance, automatic audit trails, and policy controls without tying your workloads to a single GPU provider.
Governance built in, not bolted on
An immutable audit record from design through runtime, under your own policies, enforced at runtime.
Automatic reproducibility
Exact code, environment, command, and results captured for every job. Reconstruct a run at any time.