Announcement
Webinar
Podcast
GTC 2026

CoreWeave and Meta Announce $21 Billion Expanded AI Infrastructure Agreement

Learn more
Search
Clear
Close
Close
Now Generally Available

SUNK Anywhere

SUNK Anywhere extends the industry’s first unified training system beyond CoreWeave, helping teams accelerate and scale demanding AI workloads anywhere they have infrastructure.

Get started with SUNK Anywhere
# Install SUNK
helm install sunk coreweave/sunk \
  --namespace slurm \
  -f sunk-values.yaml

# Install Slurm
helm install slurm coreweave/slurm \
  --namespace slurm \
  -f slurm-values.yaml
Copied
Why SUNK Anywhere?

Flexibility shouldn't mean fragmentation

As training environments expand across providers and customer-owned infrastructure, teams risk operating with different workflows, controls, and operating practices in each environment. That fragmentation increases complexity for both researchers and platform teams precisely where continuity matters most.

SUNK Anywhere matters because customers want a way to extend the same SUNK system beyond CoreWeave as their infrastructure footprint expands. It gives organizations a faster path to productive training outside CoreWeave, with less setup as infrastructure choices grow. Outside CoreWeave, there is still no easy way to create a research cluster, and Kubernetes-native alternatives still require significant infrastructure setup and operational assembly.

Faster path to productive training

AI teams should get to productive training faster without taking on more setup and operational overhead.

Less manual coordination

Moving workloads outside CoreWeave often introduces new setup work and manual coordination. SUNK Anywhere reduces that overhead rather than adding to it.

Infrastructure flexibility without the overhead

Platform leaders can support infrastructure outside a single provider without taking on a fragmented or brittle operating model across environments.

Less fragmentation. One unified training system

Customers are already choosing SUNK outside CoreWeave because they value the simplicity of one stack. SUNK Anywhere creates a deliberate path to broaden that adoption before fragmented alternatives define what teams expect.

sunk anywhere

The market’s most proven Slurm-on-Kubernetes offering

Architecture

One unified training system across infrastructure environments

SUNK Anywhere extends the same unified training system beyond CoreWeave, helping teams preserve consistent scheduling, workflows, and operating practices as infrastructure environments expand.

CoreWeave
‍

CoreWeave Kubernetes Service

Slurm Jobs
Kubernetes Pods
Kubernetes Pods

Cloud provider
Kubernetes

Slurm Jobs
Kubernetes Pods
Kubernetes Pods
Capabilities

Built for demanding AI workloads, wherever they run.

SUNK Anywhere extends the same unified training system customers rely on at CoreWeave to anywhere they have infrastructure — without forcing teams to adopt new tooling or different ways of operating.

More of the training environment in one system

SUNK integrates more of the training environment into one system, helping teams avoid the fragmented tooling and manual coordination that often emerge in external environments.

Hybrid scheduling and observability hooks

SUNK integrates hybrid scheduling, researcher-facing capabilities, and observability hooks, giving teams stronger visibility and control as workloads scale.

Faster path to productive training

SUNK helps teams get to a working production cluster faster, with built-in primitives for access, user environments, and containers. Teams spend less time on setup and more time on productive training work.

One way of operating across environments

SUNK Anywhere extends SUNK beyond CoreWeave rather than introducing a separate external training stack. Teams preserve familiar interfaces, scheduling behavior, and day-to-day operating practices as infrastructure footprints grow.

Left
Right

Proven by leading pioneers at production scale

See more customer stories

CoreWeave's SUNK is hands-down the industry’s leading cluster management. It's Slurm on Kubernetes and gives us the best of both. Our research scientists love using Slurm as their job orchestrator and Kubernetes gives us the observability and production-grade long-lived services that our products need. We're running large distributed jobs across thousands of GPUs on both CoreWeave and non-CoreWeave providers and deploying SUNK on our non-CoreWeave clusters required very few configuration changes.
Xander Dunn
Member of Technical Staff, Periodic Labs
Who it's for

Built for teams running demanding AI workloads across infrastructure environments.

‍

AI natives

Fast-moving AI teams that need one unified training system as their infrastructure footprint grows, without taking on fragmented tooling and added operational overhead.

  • Standardize across infrastructure environments
  • Preserve familiar Slurm workflows
  • Get to productive training faster
Copied

Enterprises

Organizations running AI across infrastructure environments that need flexibility beyond a single provider without inheriting a fragmented or brittle operating model.

  • Support multi-cloud infrastructure
  • Maintain consistency and repeatability
  • Reduce fragmentation across environments
Copied

Research labs

Research teams that need familiar interfaces and consistent scheduling behavior as infrastructure environments scale, so training work can continue without relearning how jobs are run.

  • Keep Slurm workflows intact
  • Maintain consistent scheduling behavior
  • Work without relearning how jobs are run
Copied
Left
Right
Built to deliver

What SUNK Anywhere makes possible

‍

Standardize on one unified training system

Give teams a clearer path to standardizing how demanding AI workloads are run as infrastructure footprints expand, rather than splitting training across different workflows and operating practices.

Reduce fragmentation across environments

Reduce the operational assembly that often comes with running training across multiple providers and customer-owned infrastructure, so teams do not have to stitch together as many missing pieces by hand

Simplify the path to productive AI workloads

Help platform and research teams get to productive training faster, with less setup, adaptation, and operational overhead before work can begin.

Frequently asked questions

What is SUNK Anywhere?

SUNK Anywhere extends SUNK beyond CoreWeave rather than introducing a separate external training stack. It makes the same unified training system available anywhere you have infrastructure, so teams can preserve one way of running demanding AI workloads as infrastructure environments expand.

When should teams use SUNK Anywhere?

SUNK Anywhere is designed for teams that need to run demanding AI workloads beyond a single environment without adopting different workflows, tooling, and operating practices as infrastructure changes. It is a strong fit when organizations want to support infrastructure outside CoreWeave while preserving consistency across environments.

Do teams need to change how they work to use SUNK Anywhere?

No. SUNK Anywhere is designed to preserve familiar interfaces, scheduling behavior, and day-to-day operating practices as infrastructure footprints grow. Researchers keep familiar workflows, while platform teams can standardize on one way of running demanding AI workloads across environments.

Where do sandbox workloads run?

Sandbox workloads run inside your existing CoreWeave environment, on the same infrastructure and contracted capacity already used for training.

Left
Right

Talk to us about SUNK Anywhere

See how SUNK extends outside CoreWeave and helps your team standardize on one unified training system across infrastructure environments.

Products

Solutions

AI Infrastructure

Why CoreWeave

Resources

About

© CoreWeave.
290 W Mt Pleasant Ave
Suite 4100
Livingston, NJ 07039