google_workbench_instance cost estimation
A managed Jupyter notebook VM for Vertex AI, billed the underlying Compute Engine rate per hour. An e2-standard-4 is ~$98/month if left on — idle notebooks are the waste.
A google_workbench_instance is a managed JupyterLab environment for ML development on Vertex AI. It bills the underlying Compute Engine VM rate per hour for every hour it's running — an e2-standard-4 (4 vCPU, 16 GiB) is ~$0.134/hour, about $98/month if left on around the clock. A boot disk and any attached disks bill on top.
Like every notebook environment (SageMaker, Azure ML), the defining cost problem is idle time: instances started for a session and left running for days, billing nights and weekends for an environment nobody's using. Vertex AI Workbench supports idle-shutdown, which stops the instance automatically after inactivity — the single biggest saving.
The other levers: right-size the machine type to the dev work (most exploration runs fine on an e2/n2; reserve GPU-attached instances for genuine training), and push heavy training to Vertex AI training jobs or a cluster that runs only for the job rather than a big always-on notebook.
c3x prices the instance from gce_setup.machine_type at the Compute Engine rate, so the running cost — and the cost of leaving it on — is visible up front.
Terraform example
A minimal but realistic configuration that C3X can estimate.
resource "google_workbench_instance" "dev" {
name = "ml-dev"
location = "us-central1-a"
gce_setup {
machine_type = "e2-standard-4"
disable_public_ip = true
}
}Pricing dimensions
What you actually pay for when you provision google_workbench_instance.
| Dimension | Unit | What's being charged |
|---|---|---|
| Workbench instance | per hour | The underlying Compute Engine VM rate per hour for every hour the instance runs. Stopped instances bill no compute. $0.134/hour for e2-standard-4 → ~$97.84/month if on 24/7 |
| Boot & data disks | per GB-month | Persistent disks attached to the instance bill per GB-month, including while stopped. |
Sample C3X output
An e2-standard-4 Workbench instance left running 24/7 (the avoidable case):
google_workbench_instance.dev
└─ Workbench instance (e2-standard-4) 730 hours $97.84
Monthly $97.84Optimization tips
Common ways to reduce google_workbench_instance cost without changing the workload.
Enable idle shutdown
Up to ~75% on a part-time instanceVertex AI Workbench can stop the instance automatically after an inactivity timeout — ending nights-and-weekends billing for an environment nobody's using. An instance used 8 hours a weekday costs roughly a quarter of a 24/7 one.
Right-size the machine type
Large vs an oversized GPU instanceMost ML exploration runs fine on an e2 or n2 machine. GPU-attached instances cost many times more — reserve them for work that genuinely needs the accelerator, not general notebook use.
Train as jobs, not in the notebook
Instance stays small; training is transientHeavy training on a big always-on Workbench instance pays continuously. Submitting Vertex AI training jobs (with Spot) runs the compute only for the job's duration.
Use Spot VMs for interruptible work
Up to ~91% on interruptible trainingFor fault-tolerant training, Spot VMs run 60-91% below on-demand. Keep the interactive instance small and push batch work to discounted Spot capacity.
FAQ
Does a Vertex AI Workbench instance cost money when idle?
Yes — a running instance bills the underlying Compute Engine rate whether or not you're using it. Only stopping it ends the compute charge (disks still bill). Idle running notebooks are the most common Workbench waste.
How do I stop paying for an idle Workbench instance?
Stop it when not in use and enable idle shutdown so Vertex AI stops it automatically after inactivity. That ends overnight and weekend billing — the single biggest saving on a dev notebook.
How does c3x estimate the cost?
From gce_setup.machine_type at the Compute Engine rate (shown at 24/7 as the ceiling). Real cost depends on running hours — idle shutdown and part-time use bring it well below the ceiling.
Related resources
Estimate this resource in your own Terraform
Free, open source, no API key. C3X parses your Terraform and shows line-item cost for every resource, including google_workbench_instance.