Dedicated GPU Capacity

8+ GPU whole-node or whole-cluster capacity

For customers that need long-term capacity, delivery coordination, and infrastructure-level control. Card-hour GPU usage is self-serve; 8+ GPU dedicated capacity includes dedicated support.

Contact the teamView pricing

Public status

Quote review

Workspace truth

Capacity console

Delivery mode

Commercial delivery

Evidence

Usage / billing / VaaS

What this path is for

Dedicated GPU Capacity is for customers that need 8+ GPU whole-cluster capacity, an explicit delivery plan, a capacity agreement, and stronger isolation boundaries.

  • 8+ GPU whole-node or whole-cluster capacity
  • Long-running production inference, training, or multimodal workloads
  • Delivery coordination, evaluation, acceptance, and handoff
  • Card-hour GPU usage uses the self-serve GPU rental path instead

How delivery state is determined

8+ GPU dedicated capacity does not claim instant inventory or instant activation on the public page. Status follows the capacity console, contracted scope, and delivery checkpoints.

  • Customer success team coordinates capacity planning and delivery timeline
  • Contract, invoice, budget, and acceptance records stay aligned
  • Customer-facing UI stays on the BatchIn abstraction without low-level infrastructure detail

What remains self-serve

Smaller GPU workloads, Reserved Inference, Dedicated Endpoints, and Batch remain self-serve for purchase, usage, billing, and settlement.

  • Card-hour GPU usage is paid through the capacity console
  • Every self-serve product writes workspace usage and billing ledgers
  • Only 8+ GPU dedicated capacity and Managed Deployment include dedicated delivery

Customer-facing programs

These are the public commercial states, not hardware inventory promises.

Self-serve GPU capacity

Commercial delivery

Contact the team for Dedicated GPU Capacity, especially for 8+ GPU clusters or customer-specific delivery lanes.

  • Capacity reservation and payment
  • Workspace usage and billing ledger
  • Inference, batch, and agent runtime jobs

8+ GPU dedicated capacity

Commercial delivery

Contact the team for Dedicated GPU Capacity, especially for 8+ GPU clusters or customer-specific delivery lanes.

  • Guided commercial review
  • Controlled launch and change management
  • Customer-facing BatchIn abstraction by default

Enterprise delivery

Quote review

Start with paid-pilot scoping and quote review; keep SSO, SCIM, and private connectivity human-led.

  • Quote-backed commercial review
  • Human-led delivery checkpoints
  • No public claim beyond verified contract state
Sign in to see current GPU availability by region and card type.