AWS Lets Teams Manage HyperPod AI Workspaces Directly in SageMaker Studio

AWS now lets teams create and operate SageMaker HyperPod development Spaces through Studio, with browser-based IDEs and remote VS Code access.

AWS has added a SageMaker Studio interface for creating and managing SageMaker Spaces on HyperPod clusters orchestrated with Amazon EKS. Announced on October 6, 2026, the capability lets data scientists and machine-learning engineers create, configure, start, stop, and open development environments from Studio rather than relying on command-line tools for routine operations.

From the cluster’s IDE and Notebooks tab, users can launch JupyterLab or Code Editor in a browser, or connect a local Visual Studio Code setup to a Space. The remote connection uses SSH over AWS Systems Manager Session Manager, so users do not need to manage SSH keys or expose port 22. Work in JupyterLab persists on the attached Amazon EBS volume when a Space is stopped and restarted.

The feature builds on SageMaker Spaces for HyperPod, introduced earlier in 2026, which provides interactive development environments on HyperPod EKS clusters. Administrators must first install the Spaces add-on and configure access. For browser access, AWS says the custom installation option is required; existing Studio domains may also need per-user identity propagation enabled so cluster actions are attributed to individual user profiles.

AWS says cold starts on scale-to-zero clusters typically take five to seven minutes. Its example over-provisioning setup, which keeps nodes warm and images cached, reduced startup to about 30–40 seconds. In tests using a CPU instance, a Space that fit alongside a placeholder started in about 14 seconds, while one that displaced the placeholder started in about 35 seconds. AWS notes that these figures are CPU-only and that maintaining warm nodes adds compute cost.

The add-on itself has no additional charge, according to AWS. Customers pay for the HyperPod compute their Spaces use and, for SSH-over-SSM connectivity, an hourly Systems Manager charge. Teams using over-provisioning also pay for the warm nodes they keep running.

Industry Impact

Simpler access to HyperPod workspaces

Enterprise machine learning and cloud infrastructure

Overview

SageMaker Studio now gives data scientists and ML engineers a graphical way to create and manage development Spaces on HyperPod clusters, reducing reliance on command-line operations for routine tasks. Browser-based JupyterLab and Code Editor access, plus remote Visual Studio Code connections, can make existing cluster resources easier for teams to use. Platform and security teams still need to configure the add-on and access settings. Startup speed also has cost implications: keeping nodes warm can cut waits but adds compute expense.

Key Impact Points

Teams can manage HyperPod development environments from Studio and use browser-based or local IDEs.

SSH over Systems Manager Session Manager avoids managing SSH keys or exposing port 22 for remote VS Code access.

Scale-to-zero starts typically take five to seven minutes; warm nodes can reduce startup times but incur compute costs.

Opportunities

Streamlined ML workspace operations

Enterprise machine learning and cloud infrastructure

Organizations already using HyperPod can standardize how teams provision and operate development Spaces through SageMaker Studio, giving data scientists and ML engineers a shared workflow for browser-based and local IDE access. Platform teams can pair that workflow with policies for access, persistent storage, and cluster capacity. The key operational choice is whether faster starts justify keeping nodes warm: AWS reports that warm-node configurations can reduce startup times, while noting the additional compute cost. The add-on has no extra charge, but customers pay for compute and applicable Systems Manager connectivity.

Standardized HyperPod development workspaces
Problem: Routine creation and operation of HyperPod development environments can require command-line tools, adding friction for ML teams.
Solution: Use SageMaker Studio to create, configure, start, stop, and open Spaces, with browser IDEs or remote Visual Studio Code access.
Target Users: Data scientists, ML engineers, and administrators of HyperPod clusters.
Business Value: A more consistent workspace workflow, with JupyterLab work persisting on attached EBS volumes across stops and restarts.
Revenue Potential: AWS charges for HyperPod compute and Systems Manager connectivity; keeping warm nodes also incurs compute costs. The add-on itself has no additional charge.

Reference link

https://aws.amazon.com/blogs/machine-learning/manage-amazon-sagemaker-hyperpod-spaces-directly-from-sagemaker-studio

Image Source Link

https://docs.aws.amazon.com/images/sagemaker/latest/dg/images/training-plans/tp-create-hyperpod-cluster.png

Community Discussion (0)

Loading discussions...