From bare metal to flexible virtual machines, OCI gives you a lot of ways to run a workload. We pick the shape, the tenancy, and the scaling model that fit your load and your budget, then prove it under real demand.
OCI Compute covers bare metal servers, flexible virtual machines, dedicated hosts, and GPU instances. The flexibility is useful, but it also makes over provisioning easy. We design compute around measured load, choose flexible shapes where they save money, and set autoscaling so capacity follows demand instead of sitting idle.
Because we sell no licences and carry no quota, our only goal is the shape that serves your workload at the lowest sensible cost.
We match flexible and fixed shapes to real CPU, memory, and network needs, not to a round number.
Instance pools and autoscaling configured so capacity tracks demand through peaks and quiet spells.
Dedicated bare metal for licensing or performance reasons, and GPU shapes for heavy compute.
Golden images, instance configuration, and a patch approach that keeps fleets consistent.
A quick read on when each compute model tends to fit. The right answer depends on your workload, which is what an assessment establishes.
| Compute model | Best for | Watch out for |
|---|---|---|
| Flexible VM | Most general workloads where you want to tune cores and memory | Leaving shapes oversized after launch |
| Bare metal | Licensing by physical core, or top end and consistent performance | Higher minimum cost if underused |
| Dedicated host | Isolation and licensing control with VM flexibility on top | Capacity planning sits with you |
| GPU instance | Machine learning, rendering, and heavy parallel compute | Idle GPUs are expensive, schedule them |
Book an assessment and we will show you what good looks like for your workloads, in writing, with a clear price.