How does OCI Exadata Cloud@Customer hardware differ from on-prem Exadata?
While OCI Exadata Cloud@Customer (ExaCC) and On-Premises Exadata machines may look identical from the outside—both being silver racks filled with high-performance database and storage servers—the difference lies in the ownership of the "Control Plane" and the operating model.
In 2026, Exadata Cloud@Customer X11M has become the standard for hybrid deployments, further widening the gap in how these systems are managed.
This is the most fundamental difference.
On-Premises Exadata: You buy the hardware (CapEx). You own the asset, manage the maintenance contracts, and pay for all the CPU cores upfront via perpetual licenses.
Exadata Cloud@Customer: Oracle owns and leases the hardware to you (OpEx). You pay a monthly base floor fee for the infrastructure and then pay-as-you-go for the OCPUs you actually use. It is a "Cloud" billing model sitting in your data center.
The physical architecture of ExaCC is designed to create a hard wall between Oracle's infrastructure management and your data.
| Task | On-Premises Exadata | Exadata Cloud@Customer |
| Firmware/BIOS Updates | Customer (Manual) | Oracle (Automated/Remote) |
| Hypervisor (Dom0) Patching | Customer (Manual) | Oracle (Automated/Remote) |
| Storage Cell Updates | Customer (Manual) | Oracle (Automated/Remote) |
| Database/OS Patching | Customer (Full Control) | Customer (via OCI Console/API) |
| Physical Hardware Repair | Customer (Coordinates) | Oracle (Monitors & dispatches) |
Pro-Tip for the Blog: In ExaCC, Oracle manages the "Infrastructure" (the racks, switches, and storage cells), while you retain full
rootaccess to the Virtual Machines (DomU) where your databases live.
In 2026, the scaling of these two systems follows very different physical paths:
On-Premises Scaling: If you need more power, you purchase more "Elastic Expansion" kits (physical servers). This involves a procurement cycle, shipping, and manual installation by your team.
ExaCC Scaling (Cloud Elasticity): The physical rack is often "over-provisioned" with more cores than you initially use. You can scale from 2 OCPUs to hundreds of OCPUs in minutes via the OCI Console. The hardware is already there; you just "turn it on" digitally.
ExaCC is physically connected back to an OCI Region (via a highly secure, encrypted metadata tunnel). This allows it to use Cloud Automation that on-prem machines lack:
Autonomous Database: You can run Oracle Autonomous Database on ExaCC hardware. This is physically impossible on standard on-prem Exadata because Autonomous requires the OCI control plane to handle the "self-driving" logic.
One-Click Lifecycle: Tasks like creating a Data Guard standby, enabling backups to OCI Object Storage, or performing a rolling patch of a 10-node cluster are "one-click" operations in ExaCC. On-prem, these are complex manual projects.
On-Prem: You are responsible for the entire network stack, including the spine switches and integration into your corporate core.
ExaCC: Oracle provides the Top-of-Rack (ToR) switches as part of the managed service. The internal "east-west" traffic (RoCE/InfiniBand) is completely managed by Oracle. You only provide the "uplinks" to your data center network.
"Choose On-Premises Exadata if you want total 'dark site' air-gapped control and have a large team of DBAs to manage the hardware. Choose Exadata Cloud@Customer if you want the 'Self-Driving' experience of the cloud but your data is legally required to stay behind your own firewall."