What is BYOC for AI training?

BYOC — bring your own cloud — for AI training means you run training jobs on compute you already own and pay for, while a separate control plane adds the management and reliability layer. You keep your cloud relationship; the platform makes it dependable.

BYOC, defined

In a BYOC model, the platform never owns the compute. Your jobs run on your cloud account, reserved instances, credits, or a committed GPU contract — and are billed by your provider, not the platform. The platform's job is everything around the GPU: provisioning, orchestration, checkpoint storage, and recovery.

Why teams choose BYOC

• Use existing commitments — put cloud credits, reserved instances, and contracted GPUs to work.
• Keep your data and pricing — compute and storage stay in your account under your terms.
• Avoid a per-run markup — with VaultLayer, BYOC runs carry no per-run GPU charge.

BYOC with VaultLayer

VaultLayer is BYOC-first. Connect a cloud with vl connect, then run vl run --byoc python train.py; the job provisions on your account, checkpoints to your bucket, and auto-resumes on failure. It fails closed if no credentials are set and never routes a BYOC job to a managed pool. See the BYOC GPU training overview.

Frequently asked questions