Run AI training on your own GCP GPUs
VaultLayer's BYOC model runs your training jobs on your own Google Cloud account — your GPU instances, your bucket, your pricing — while adding orchestration, checkpointing, and recovery on top. Connect once and run your existing script.
Connect your GCP account
vl connect gcp # service account JSON + project IDVaultLayer uses a service account scoped to your project to provision GPUs. Credentials are used only to run your jobs and never printed in logs.
Run on your GCP GPUs
vl run --byoc train.pyThe job provisions on your Google Cloud project, checkpoints to your bucket, and auto-resumes from the last step if an instance is interrupted. A BYOC job never routes to a managed pool and fails closed if no credentials are set. Because the compute is billed by Google Cloud under your account, there is no per-run charge from VaultLayer.
Why run on your own GCP
Keep your data, pricing, and account — and put existing GCP commitments and credits to work. See using your cloud credits and reserved GPUs and the BYOC overview. AWS teams: run training on your own AWS GPUs. BYOC onboarding is a high-touch beta — talk to us to get set up.
Frequently asked questions
Does my training data leave my GCP project?
No. Compute and storage stay in your own Google Cloud project — jobs run on your instances and checkpoint to your own bucket. VaultLayer adds the control plane, not the compute.
What do I need to connect GCP?
A service account JSON and your project ID, set via vl connect gcp. The service account is scoped to your project and used only to run your jobs.
Keep every training job moving.
Sign up, install the CLI, and submit your first training job in minutes — on your own cloud or elastic GPU capacity.
Sign upEvaluating for a team, or bringing your own cloud or GPU contract? Talk to us — you'll reach the founder directly.