Bringing existing Oracle licences to AWS or Azure is one of the most common workload moves in the enterprise, and one of the most consistently mis-licensed. The Oracle Cloud Licensing Policy translates Processor licences differently for authorised clouds, applies different rules to RDS for Oracle than to EC2, and reserves the right to change the policy unilaterally. Here is what the BYOL rules actually require.
Oracle's "Licensing Oracle Software in the Cloud Computing Environment" document — referred to informally as the Cloud Licensing Policy — is the authoritative source for how on-premises Oracle licences translate to deployment on AWS, Azure, and (with a recent amendment) Google Cloud. Importantly, the document is a policy and not a clause in the customer's contract. Oracle can update it. Recent updates have tightened the vCPU translation on some instance families and removed GCP, AWS, and Azure from "Authorized Cloud Environments" in a brief 2024 revision, with that change subsequently softened. The current rule should be re-verified at every renewal.
Under the current policy, the vCPU-to-Processor translation on AWS and Azure is: where hyperthreading is enabled, every two vCPUs count as one Oracle Processor licence. Where hyperthreading is not enabled (or where the customer cannot demonstrate it), one vCPU counts as one Processor. The translation on OCI is more favourable to Oracle as a vendor — one OCPU equals one Processor, an OCPU being equivalent to one physical core. The translation on GCP follows the AWS/Azure rule.
NUP licensing translates to cloud with the same 25 per Processor minimum and the same definitional rules as on-premises. The Processor count is calculated using the cloud vCPU translation rather than the Core Factor Table. A 4-vCPU AWS instance running Oracle EE under NUP requires a minimum of 50 NUPs (2 Processors × 25 NUP/Processor).
The vCPU translation, the policy version in force, and the contractual fall-back should all be modelled before commitment.
Amazon RDS for Oracle and Azure SQL Database for Oracle Database (and equivalent Google Cloud SQL deployments) are managed database services that run Oracle under licence-included or BYOL options. The licensing rules differ from EC2/Azure VM/GCE BYOL deployment.
Under RDS BYOL, the customer supplies the Oracle licence and AWS supplies the managed infrastructure. The vCPU translation rule applies — every two vCPUs is one Processor with hyperthreading. RDS Multi-AZ failover replicas require their own licensing under the standard rules; the "passive standby" exemption available on-premises does not apply.
License-Included RDS bundles the Oracle Database SE2 licence with the infrastructure. Enterprise Edition is not available license-included. License-Included RDS is bounded to SE2 instance types and is a useful option for small workloads where SE2 features are sufficient.
Oracle Options (Partitioning, Advanced Compression, etc.) under RDS BYOL must be separately licensed by the customer. RDS itself does not provision Option licences. The same is true on Azure and GCP managed Oracle services.
The BYOL rules with worked examples for AWS, Azure and GCP.
Oracle Cloud Infrastructure has structurally favourable BYOL economics — one OCPU per Processor licence, no vCPU translation discount, Oracle-authorised by default, and several programmes (Universal Credits, ULA compatibility) that reduce friction. For Oracle workloads where the choice of cloud is open, the economic case for OCI is straightforward. Where the cloud choice is constrained — multi-cloud strategy, application affinity to AWS/Azure, regulatory boundaries — OCI is one option among several, and the BYOL economics differ accordingly.
The right cloud is the one where the workload runs best — but the licence math should be in the comparison.
For any Oracle workload migration with material cloud spend, the independent BYOL model is the single highest-leverage piece of advisory work in the project — and the natural entry point to broader Oracle license optimization once the estate is sized correctly for the target platform.
Our cloud and Oracle practices build defensible BYOL deployment models for buyers.
Weekly compliance intelligence for IT leaders.