A Google Colab alternative that doesn't erase your work
Google Colab is great for quick experiments — until the session dies, your files vanish and your environment resets. NotebookHub gives you a real, persistent workspace: JupyterLab or VS Code in your browser, with your files, packages and data exactly where you left them.
Why people leave Colab
Colab notebooks run on ephemeral virtual machines. Close the tab or hit an idle timeout and the runtime is recycled: installed packages are gone, uploaded data is gone, and long-running jobs are killed mid-training.
- Sessions disconnect after idle timeouts — long jobs die silently
- The filesystem is wiped between sessions; everything must be re-uploaded or re-mounted
- pip installs run again on every session start
- Resources are shared and variable — one day fast, next day throttled
What NotebookHub does differently
Every workspace is a dedicated container on our Kubernetes infrastructure, with guaranteed CPU and RAM (Kubernetes QoS 'Guaranteed' — your resources are reserved, not shared) and a persistent disk that survives pauses and restarts.
- Persistent storage: your notebooks, data and installed packages stay put
- JupyterLab AND VS Code — same workspace, your choice of editor
- Dedicated resources per workspace (from 1 vCPU / 8 GB RAM on the free plan)
- Pause a workspace and resume days later exactly where you left off
- Kernels for Python, R, geospatial and PyTorch work
Where Colab still wins
Honesty first: Colab offers free GPU time, and NotebookHub does not sell GPU workspaces yet (it's on the roadmap). If your workload is GPU-bound training, Colab or a GPU cloud is the right tool today. If your pain is lost sessions, lost files and unreproducible environments for CPU work — that is exactly what NotebookHub fixes.
Frequently asked questions
Is there really a free plan?
Yes — one persistent workspace with 1 vCPU, 8 GB RAM and 50 GB of storage, JupyterLab or VS Code, no credit card required.
Can I import my Colab notebooks?
Yes. Download your .ipynb from Colab and upload it in JupyterLab — it's standard Jupyter, so everything opens as-is.
Do my packages survive a restart?
Yes. The workspace disk is persistent: pip/conda installs, data and outputs are still there when you come back.
Where does my code run?
On our own Kubernetes infrastructure in Europe, in an isolated container per workspace. Your files are never shared with other users.
Stop re-uploading your data every morning
One persistent workspace free, forever. No credit card.