How Much Does Oracle Cloud at Customer Cost
Oracle Cloud@Customer brings Oracle’s cloud services into your data center with a subscription model.
In 2025, costs are structured in two parts: a fixed infrastructure fee for the on-prem hardware and a consumption-based charge for the cloud resources (measured in Oracle CPU hours, storage GBs, etc.).
Below, we break down current pricing for the main Cloud@Customer offerings – Exadata Cloud@Customer, Compute Cloud@Customer, and OCI Dedicated Region – including how Bring Your Own License (BYOL) versus license-included options affect the bill.
We also clarify how Oracle’s Universal Cloud Credits apply and provide realistic monthly cost examples at different scales.
Pricing Components in 2025
Oracle Cloud@Customer uses a hybrid pricing model: you pay a fixed monthly subscription for the dedicated on-prem hardware (with a multi-year term commitment), plus usage fees for the cloud services consumed on that hardware.
Key cost components include:
- Infrastructure Base Fee: A flat monthly Cloud@Customer hardware and management charge. (Requires a 4-year term for Cloud@Customer deployments.)
- Compute (OCPU/ECPU) Usage: Measured per Oracle CPU core hour (called OCPU or ECPU on Exadata). This is billed similarly to Oracle’s public cloud rates.
- Memory and Storage Usage: Measured in GB per month (for storage) or per hour (for allocated memory in Compute). Storage comes in tiers (e.g., block storage, object storage) with their own $/GB-month rates.
- Support and Management: Included in the subscription fee for license-included services, but if using BYOL, you continue to pay support for your own licenses separately.
- Oracle Licensing: You can choose either BYOL (bring your own license)—you supply Oracle software licenses—or License-Included—the cloud subscription includes the Oracle software license. This choice impacts your cost per OCPU.
Universal Cloud Credits come into play for the usage part of Cloud@Customer. Consumption of OCPUs, storage, etc., is metered like Oracle’s public OCI services, drawing down from your Universal Credit balance (if you have a committed spend contract).
However, universal credits do not cover the infrastructure base fees (they are a fixed subscription separate from consumption).
An enterprise signs a contract for a Cloud@Customer deployment (covering hardware and a minimum spend). Then all usage charges (OCPU hours, etc.) are either billed monthly (pay-as-you-go) or deducted from their committed universal credits.
Exadata Cloud@Customer Pricing (2025)
Oracle Exadata Cloud@Customer is a database-optimized Cloud@Customer offering. Pricing involves a monthly infrastructure fee based on the Exadata system size plus database OCPU usage fees (at different rates for BYOL vs. license-included).
- Exadata Infrastructure Subscription: As of 2025, Oracle offers multiple Exadata Cloud@Customer sizes. For example, a Base System (X10M generation) with two database servers and three storage servers costs about $8,000 monthly for the hardware. Larger configurations scale up to a Full Rack (X10M) at roughly $43,200 per month for infrastructure. Newer X11M generation systems have adjusted pricing – e.g., an X11M base system is around $3,200 for the rack plus required storage servers (~$4,800), totaling about $8,000/month. These fees cover the Exadata hardware on-premises, Oracle management, and support for the hardware over a 4-year term.
- OCPU/ECPU Usage Charges: On Exadata, you pay for actual database compute usage by the hour. An OCPU is one Oracle CPU core (with hyperthreading, equals two vCPUs). For license-included, the list price is about $1.344 per OCPU-hour (or ~$0.672 with an annual universal credit commitment) for Exadata database usage. For BYOL, the rate is much lower – around $0.3226 per OCPU-hour (or ~$0.1613 with commit). BYOL assumes you already own Oracle Database licenses and pay support for them; thus, Oracle only charges for the infrastructure portion. License-included pricing is roughly 4X higher per OCPU because it bundles in the Oracle Database license and support costs. Storage usage on Exadata (for database data, backups, etc.) is charged per TB-month (e.g., Oracle’s block storage ~$0.0425 per GB-month for high performance).
- Minimums and Scaling: Oracle requires a minimum activation of 8 OCPUs (or ECPUs) per database node on Exadata Cloud@Customer. However, you can scale to zero OCPUs on a VM when not in use, stopping billing for those cores. In other words, if you shut down all database VMs, you pay no OCPU usage fees (only the base infrastructure continues). This allows cost savings for idle periods, although production databases will run continuously in practice. Still, non-production environments can be powered off to avoid charges. Oracle’s billing is granular to the second after a minimum 1-minute or 48-hour initial period for new instances.
- BYOL vs License-Included Impact: If you bring your own licenses, ensure they cover the same editions/options you use on Exadata (e.g., Enterprise Edition, Multitenant, RAC, etc.). BYOL OCPU usage is ~75% cheaper, but you carry the separate annual support contract costs for those licenses. License-included is simpler (Oracle handles all licensing), but at a premium cost. Example: A small Exadata Cloud@Customer deployment with 16 OCPUs active on average:
- BYOL case: Infrastructure ~$8,000 + 16 OCPUs * 720 hrs * $0.3226 = ~$8,000 + $3,720 = $11,720 per month (plus whatever you pay Oracle annually for support on the licenses you brought).
- License-Included case: Infrastructure ~$8,000 + 16 * 720 * $1.344 = ~$8,000 + $15,514 = $23,514 per month, all-inclusive.
Assuming you already own the licenses, this illustrates ~50% savings using BYOL in this scenario.
Read Oracle Dedicated Region Cost Optimization.
Compute Cloud@Customer Pricing (2025)
Oracle Compute Cloud@Customer extends OCI’s core compute and storage services on-prem. Its pricing model mirrors Oracle’s public cloud pricing, with a smaller infrastructure base fee than Exadata.
- Compute Infrastructure Subscription: Oracle provides a managed rack with server and storage hardware. A typical Compute Cloud@Customer base system (using Intel E5 CPUs, supporting up to 552 OCPUs and ~6.7 TB RAM) costs about $4,900 monthly for the base hardware. This is a 4-year term subscription for the rack. If you require additional storage capacity beyond the base, storage expansion subscriptions are available (e.g., a “Balanced Storage Expansion” supporting ~175 TB is ~$800 per month, and a “Performance Storage” expansion (flash-heavy) is ~$1,900 per month). There is also a GPU configuration (with NVIDIA GPUs in the rack) at a higher base cost (the largest GPU-enabled config was priced around $30,000/month in Oracle’s list for a 4-GPU setup). The Compute Cloud@Customer subscription includes support and maintenance of the on-prem equipment.
- Compute and Storage Usage Charges: Running VMs or bare-metal instances on Compute Cloud@Customer incurs usage fees identical to OCI public rates. For instance, Standard VM OCPUs are about $0.03 per OCPU-hour (pay-as-you-go), and memory is ~$0.002 per GB-hour. A VM with 4 OCPUs and 30 GB RAM running full-time would cost ~($0.034 + $0.00230) * 720 = $86.4 per month. Block storage is charged around $0.0255 per GB-month for standard volumes, so 10 TB of attached storage would be ~$261 monthly. Object storage in Cloud@Customer is the same $0.0255/GB-month as OCI public. Data egress to the internet, if any, is charged at the same network rates as OCI. Notably, Oracle doesn’t charge extra for on-prem data ingress or egress between your data center and the Cloud@Customer itself (since it’s local).
- Examples: Because Compute Cloud@Customer base hardware supports a lot of capacity, the cost per resource can be very low at high utilization. For example, consider a moderate usage scenario: you use 100 OCPUs and 1.2 TB of RAM on average, plus 50 TB of block storage:
- Base fee: $4,900 (hardware).
- Compute usage: 100 OCPUs * 720 hrs * $0.03 = $2,160; 1,200 GB RAM * 720 hrs * $0.002 = $1,728.
- Storage: 50,000 GB * $0.0255 = $1,275 per month.
- Total: approx $10,063 per month.
If utilization were doubled (200 OCPUs, etc.), the usage costs would double, but the base would stay the same, effectively lowering the average unit cost. Conversely, if the rack is underutilized (only a few dozen OCPUs are used), the fixed $4.9k looms large. At minimal usage (say 10 OCPUs, 100 GB RAM, 10 TB storage), you might pay ~$4,900 + $216 + $144 + $255 = ~$5,515, meaning the majority is the base fee.
- Licensing: Compute Cloud@Customer itself is infrastructure – you can run any software on the VMs, so there’s no Oracle license automatically included except the OS (Oracle Linux), which is free. Suppose you run Oracle Database or other Oracle software on these VMs. In that case, you can apply BYOL or consume license-included cloud services (like an Autonomous Database on the Compute C@C via OCI console). BYOL is common, e.g., you can deploy your Oracle Middleware or DB and allocate your existing licenses to those VMs. Oracle treats Compute Cloud@Customer just like OCI public regarding licensing rules: you must license the OCPUs of any Oracle software you run (with the same core factor/counting as in OCI). The benefit is that you can turn resources on/off to match license use.
Dedicated Region Cloud@Customer Pricing (2025)
OCI Dedicated Region Cloud@Customer provides an entire isolated Oracle Cloud region with all OCI services on your premises.
It has a unique pricing structure: instead of specific hardware fees, Oracle requires a minimum annual spending commitment for the region, and all services inside are charged at normal OCI consumption rates against that commitment.
- Minimum Commitment: Oracle dramatically lowered the entry cost for the Dedicated Region. Initially launched at $6 million per year (with a 3-year minimum) in 2020, it was reduced in 2022 to about $1 million per year with a 4-5 year minimum commitment. As of 2025, Oracle indicates you can “start small” with a five-year commitment, meaning roughly $5 million over 5 years as a baseline. This comes with a smaller footprint (about 12 racks of gear to start, instead of ~50 racks previously). The key idea is you commit to spending a certain amount on cloud services per month/year, and Oracle deploys enough hardware to support that, scaling as needed.
- Service Usage Charges: Within a Dedicated Region, service pricing is exactly the same as in Oracle’s public cloud. You pay for what you use (compute, DB, storage, etc.) at the same list prices, and it draws down against your committed credits. You pay overages at the same rates if you consume beyond your committed amount. All the OCI services – from VMs to Autonomous Database to Analytics – are available. Universal Credits are inherently the model here: your commitment is a purchase of a large block of universal cloud credits to be used in your on-prem region. Oracle still provides the hardware, management, and upgrades as part of the deal (no separate “hardware fee” line item; that is baked into the fact that you must commit to a certain spend).
- Example: Suppose a company signs on for a Dedicated Region with a $1M/year commitment (approximately $83K per month). The organization can then use any OCI services in that region up to $83K worth of monthly consumption (or average – the usage might fluctuate, typically the commitment is annual). For instance, $83K could cover roughly 2,500 OCPUs of compute running full-time, or a mix of VMs, databases, storage, etc. If, in one month, they only use $50K worth, they will still pay $83K (the minimum). If they consistently need more capacity, they would arrange to increase the committed amount (the hardware cluster can scale out with more racks). Oracle’s approach is to right-size the installed hardware to your committed needs and then grow it as your usage grows, with Oracle managing hardware additions transparently.
- Comparing Cost to Public Cloud: Dedicated Region is economically viable for large enterprises. The commit ensures a predictable high spend, but you gain data residency and isolation. For companies already spending millions on OCI, bringing it on-prem might not increase cost – you’re just moving where the money is spent. Oracle also touts that there are no additional premiums on the service pricing for being on-prem; the unit prices remain standard. The trade-off is the commitment term: you’re effectively locked into a significant Oracle spend for 5 years. For organizations that can commit, this can be cheaper than building an equivalent private cloud, especially considering it includes management and updates.
Monthly Cost Examples at Different Scales
To tie it all together, here are a few scenarios with ballpark monthly costs:
- Small Deployment (Entry Exadata): A bank needs an on-prem cloud for a few critical databases. They chose an Exadata Cloud@Customer Base System. They BYOL for Oracle Database (Enterprise Edition with options). They keep 8 OCPUs active for production, occasionally bursting to 16 for backups. The approximate cost is $8,000 (base) + $2,000 usage = $10,000/month. (If they had no licenses and used license-included, the same usage would cost ~$8,000 + $6,000 = $14,000/month.)
- Mid-size Hybrid Use (Compute @ Customer): A tech company has latency-sensitive app servers and 100 TB of data that must stay on-prem, but wants cloud flexibility. They get a Compute Cloud@Customer. Over a month, they average 200 OCPUs of compute across various VMs, 4 TB of RAM, and 100 TB of storage, and use Oracle Database via BYOL on some VMs. Approx. cost: $4,900 (base) + $4,320 (CPU) + $5,760 (RAM) + $2,550 (storage) = $17,530/month. This supports dozens of VMs and large storage. In the OCI public region, the pure usage cost would be similar ($12,630 in this case), but the Cloud@Customer adds the fixed $4.9k. They pay that premium for on-prem control.
- Large-Scale Dedicated Region: A government signs up for a Dedicated Region with a $1.2M/year commitment (100K/month). They utilize it fully, running a mix of services: 100 OCPUs on web servers, 50 OCPUs on databases (BYOL), 200 TB of various storage, plus some analytics jobs. Rough breakdown of consumption: $40K on compute, $20K on database, $20K on storage, $20K on various PaaS services = $ 100 K. Cost: They pay a steady $100K per month per the commitment. If they under-utilize, they still pay $100K (with room to grow). If they over-utilize (say $120K worth one month), the excess $20K is billed additionally or carried against any annual buffer.
These examples illustrate that Oracle Cloud@Customer can cost from five figures per month for a small footprint to six figures per month for a dedicated on-prem region.
BYOL vs. included licensing can dramatically swing the costs, and high utilization of the provided hardware yields the best cost efficiency. Conversely, deploying Cloud@Customer for very light workloads would be cost-inefficient due to the base fees. It’s best suited for sizable, steady workloads where data residency or latency is paramount.
Bottom Line: In 2025, Oracle Cloud@Customer pricing will become more transparent and aligned with cloud consumption models. Oracle has lowered entry points (e.g., smaller Exadata configs, smaller Dedicated Region footprint) to make it accessible to more customers. By understanding the components—base subscription, OCPU usage, storage, and license model—you can estimate realistic costs and tailor the model (BYOL or included) to your advantage.
