About Docker Offload

Table of contents

Docker Offload is a fully managed service for building and running containers in the cloud using the Docker tools you already know, including Docker Desktop, the Docker CLI, and Docker Compose. It extends your local development workflow into a scalable, cloud-powered environment, so you can offload compute-heavy tasks, accelerate builds, and securely manage container workloads across the software lifecycle.

Docker Offload also supports GPU-accelerated instances, allowing you to containerize and run compute-intensive workloads such as Docker Model Runner and other machine learning or data processing tasks that benefit from GPU.

Key features

Docker Offload includes the following capabilities to support modern container workflows:

  • Cloud-based builds: Execute builds on remote, fully managed BuildKit instances
  • GPU acceleration: Use NVIDIA L4 GPU-backed environments for machine learning, media processing, and other compute-intensive workloads.
  • Ephemeral cloud runners: Automatically provision and tear down cloud environments for each container session.
  • Shared build cache: Speed up build times across machines and teammates with a smart, shared cache layer.
  • Hybrid workflows: Seamlessly transition between local and remote execution using Docker Desktop or CLI.
  • Secure communication: Use encrypted tunnels between Docker Desktop and cloud environments with support for secure secrets and image pulling.
  • Port forwarding and bind mounts: Retain a local development experience even when running containers in the cloud.
  • VDI-friendly: Use Docker Offload in virtual desktop environments or systems that don't support nested virtualization.

Why use Docker Offload?

Docker Offload is designed to support modern development teams working across local and cloud environments. It helps you:

  • Offload heavy builds and runs to fast, scalable infrastructure
  • Accelerate feedback loops in development and testing
  • Run containers that require more resources than your local setup can provide
  • Build and run AI apps with instant access to GPU-powered environments
  • Use Docker Compose to manage complex, multi-service apps that need cloud resources
  • Maintain consistent environments without managing custom infrastructure
  • Develop efficiently in restricted or low-powered environments like VDIs

Docker Offload is ideal for high-velocity development workflows that need the flexibility of the cloud without sacrificing the simplicity of local tools.

How Docker Offload works

Docker Offload replaces the need to build or run containers locally by connecting Docker Desktop to secure, dedicated cloud resources.

Building with Docker Offload

When you use Docker Offload for builds, the docker buildx build command sends the build request to a remote BuildKit instance in the cloud, instead of executing it locally. Your workflow stays the same, only the execution environment changes.

The build runs on infrastructure provisioned and managed by Docker:

  • Each cloud builder is an isolated Amazon EC2 instance with its own EBS volume
  • Remote builders use a shared cache to speed up builds across machines and teammates
  • Build results are encrypted in transit and sent to your specified destination (such as a registry or local image store)

Docker Offload manages the lifecycle of builders automatically. There's no need to provision or maintain infrastructure.

Note

Docker Offload builders are currently hosted in the United States East region. Users in other regions may experience increased latency.

Running containers with Docker Offload

When you use Docker Offload to run containers, a Docker Desktop creates a secure SSH tunnel to a Docker daemon running in the cloud. Your containers are started and managed entirely in that remote environment.

Here's what happens:

  1. Docker Desktop connects to the cloud and triggers container creation.
  2. Docker Offload pulls the required images and starts containers in the cloud.
  3. The connection stays open while the containers run.
  4. When the containers stop running, the environment shuts down and is cleaned up automatically.

This setup avoids the overhead of running containers locally and enables fast, reliable containers even on low-powered machines, including machines that do not support nested virtualization. This makes Docker Offload ideal for developers using environments such as virtual desktops, cloud-hosted development machines, or older hardware.

Docker Offload also supports GPU-accelerated workloads. Containers that require GPU access can run on cloud instances provisioned with NVIDIA L4 GPUs for efficient AI inferencing, media processing, and general-purpose GPU acceleration. This enables compute-heavy workflows such as model evaluation, image processing, and hardware-accelerated CI tests to run seamlessly in the cloud.

Despite running remotely, features like bind mounts and port forwarding continue to work seamlessly, providing a local-like experience from within Docker Desktop and the CLI.

Docker Offload provisions an ephemeral cloud environment for each session. The environment remains active while you are interacting with Docker Desktop or actively using containers. If no activity is detected for about 5 minutes, the session shuts down automatically. This includes any containers, images, or volumes in that environment, which are deleted when the session ends.

What's next

Get hands-on with Docker Offload by following the Docker Offload quickstart.