Flexible workloads
Development, evaluation, bursty inference and project-based fine-tuning that should not lock an entire dedicated server.
GPU Instances
Flexible GPU Instances for Enterprise AI
Virtualised or fractional GPU access when the workload needs elasticity more than exclusive physical servers.
Overview
What are GPU instances?
GPU instances provide virtualised or fractional access to accelerated compute. They are the right model when teams need GPU capacity that can flex with projects, experiments and production services, without reserving an entire physical server for every job.
Dimension AI delivers GPU instances as one of three GPUaaS consumption models, alongside Bare Metal GPU and GPU clusters, all operating inside the same AI Factory architecture.
Capabilities
Where GPU instances fit
Right-sized access
Fractional GPU consumption when a full node is more capacity than the application needs.
Path to dedicated or cluster scale
Start on instances, then move to Bare Metal GPU or GPU clusters as isolation or multi-node performance becomes the constraint.
Capabilities
Still an enterprise GPU platform
Factory operations
Instances are scheduled and observed inside Dimension AI orchestration, not offered as unmanaged virtual machines.
Connected services
The same environment can expose inference APIs and model endpoints when the workload is ready to be consumed as a service.
FAQ
Questions teams ask
When should I use GPU instances instead of Bare Metal GPU?
Use instances when you need flexible or fractional access. Use Bare Metal GPU when the workload requires exclusive physical servers and full control of the machine.
