If you landed here looking for how to sign up for AWS SageMaker Studio Lab, there's a change worth knowing before you go further: AWS's documentation now states plainly that Studio Lab is no longer open to new customers. Existing Studio Lab users can keep using their accounts, and AWS says it will continue security and availability updates, but it will not add new features and it will not let new people sign up.
That matters because Studio Lab's whole pitch, going back to its 2021 launch, was a genuinely free, no-credit-card, no-AWS-account notebook environment aimed at students and beginners. That on-ramp is gone for anyone new. This article covers what replaced it and how to actually get started with machine learning on AWS, or elsewhere, right now.
What Happened to SageMaker Studio Lab
AWS frames the shutdown as a consolidation, not an abandonment. Per their Studio Lab availability change notice, Amazon SageMaker Studio is positioned as "a complete replacement for SageMaker Studio Lab," offering the same JupyterLab-based notebook runtime, the same conda environments, and the same file management workflow Studio Lab users were used to.
The difference is access. Studio Lab let you start without an AWS account at all; just an email address and a request for access. SageMaker Studio requires:
- A full Amazon Web Services account (studio.lab.aws is being folded into the main AWS ecosystem)
- Setting up a SageMaker Studio domain and user profile
- Creating a JupyterLab Space, each with its own Amazon EBS volume for persistent storage
AWS also quietly updated Studio Lab's underlying notebook runtime to JupyterLab 4 (as of August 8, 2025) for the customers who still have access, so if you're one of the existing users, note that some previously installed extensions may need reinstalling due to the dependency shift.
Why This Matters If You're New to Machine Learning
The original appeal of Studio Lab was removing every barrier between "I want to learn ML" and "I'm running a notebook." No credit card, no cloud account, no billing risk. That specific on-ramp doesn't exist inside AWS anymore for new users.
This doesn't mean AWS has stopped being a reasonable place to learn. It means the entry point changed, and skipping that detail would have you trying to sign up for a service that no longer accepts new accounts.
What to Use Instead in 2026
If you want to stay AWS-native: Amazon SageMaker Studio is the direct successor and the only path AWS points to. It requires an AWS account and a configured domain, but AWS's free tier still applies to a meaningful amount of SageMaker compute, and SageMaker Studio supports the same JupyterLab notebooks, conda environments, and ML tooling Studio Lab had. It also adds things Studio Lab never had: built-in SageMaker training jobs, model registries, and deeper integration with other AWS services, which matters once a project grows past a single notebook.
Best fit: Teams or learners who already use AWS for other infrastructure and want their ML work to live in the same account and billing structure.
Ask about: Whether your use case needs an AWS account tied to a company billing setup, since SageMaker Studio no longer offers the zero-commitment signup Studio Lab had.
If you want a true no-account free start: Google Colab remains the closest match to what Studio Lab originally offered: open a notebook in a browser, run Python with GPU/TPU access, no AWS account needed. Kaggle Notebooks is another genuinely free option with a large library of public datasets and community notebooks attached, useful if you're learning by examining other people's work.
Best fit: Students, hobbyists, or anyone who wants to start writing and running ML code today without creating a cloud account.
Ask about: Session time limits and compute availability, since free tiers on Colab and Kaggle both cap GPU time and can deprioritize free users during high demand.
How to Choose a Starting Environment
- Do you already have, or plan to need, an AWS account for other infrastructure? If yes, SageMaker Studio keeps everything in one place. If no, setting up a full AWS account just to try machine learning is a heavier first step than it needs to be.
- Is this a one-off learning exercise or the start of a real project? Colab and Kaggle are fine for coursework, experimentation, and portfolio projects. Once you need scheduled training jobs, model deployment, or integration with production data pipelines, SageMaker Studio's deeper AWS tooling becomes worth the setup overhead.
- How much do you care about GPU access and limits? Free-tier GPU access on Colab and Kaggle is real but capped and can be deprioritized under load. SageMaker Studio's compute is pay-as-you-go beyond the AWS free tier, which is more predictable if you need guaranteed capacity.
- Do you need production-grade MLOps down the line? If the end goal is deploying a model into a real product, starting in SageMaker Studio (or at minimum, treating it as the eventual destination) avoids a migration step later.
If your team is evaluating AWS for a production machine learning workload, rather than a beginner tutorial, that's a different conversation: one about data pipelines, model governance, and deployment architecture. Our AI consulting services cover that scoping work if you're past the "which free notebook do I use" stage.
