A Reference Engine That Was Always Out of Reach
Exadata has long held a singular position in the enterprise database landscape. Its performance on demanding workloads — high-frequency OLTP, large-scale data warehouses, multi-node RAC clusters — is hard to argue with. But the traditional model came with a heavy upfront commitment: two database servers, three storage servers, dedicated infrastructure. The bill was running into the millions before a single query ever executed.
Since August 13, 2026, that logic has changed. Oracle made its Exadata Database on Exascale Infrastructure service (ExaDB-XS) generally available on AWS as part of Oracle AI Database@AWS. The key shift: a fully consumption-based model with no minimum infrastructure commitment.
What the On-Demand Model Actually Changes
The break is architectural first. ExaDB-XS is built on a shared storage pool, fully decoupled from compute nodes. In practice, an organization can start with 8 ECPUs (Elastic CPUs) and 300 GB of storage, then scale each resource independently based on actual load — from the AWS console, API, or CLI. There are no separate IOPS charges: only compute consumed and storage occupied appear on the bill.
Oracle claims an approximately 95% reduction in entry cost compared to a traditional dedicated Exadata system. For data teams, this unlocks use cases that were previously economically unworkable:
- Development and CI/CD environments: Instant thin cloning (redirect-on-write technology) creates database copies without duplicating storage. A clone only consumes additional space once its data diverges from the source.
- Multi-region disaster recovery: The service is available across 22 AWS regions worldwide (United States, Europe, Asia-Pacific, Latin America, Australia), making it viable to deploy a cross-region standby at a cost calibrated to actual usage.
- On-demand analytical workloads: Teams can provision compute power for an intensive analysis window, then scale back down with no penalty.
AWS-Native Integration — No Compromises Required
Unlike hybrid approaches that force teams to juggle two separate consoles, ExaDB-XS integrates natively with standard AWS tooling. Backups go to Amazon S3, encryption flows through AWS KMS with customer-managed keys, and metrics surface in CloudWatch and EventBridge. IAM, CloudTrail, Secrets Manager, Transit Gateway, VPC Lattice, and Cloud WAN are all natively supported.
The service supports Oracle Database 19c and 26ai — the latter including vector search via AI Smart Scan, processed directly at the storage layer.
A Deepening Oracle-AWS Partnership
ExaDB-XS is part of a broader strategic shift. Oracle AI Database@AWS launched in July 2025. In April 2026, Oracle and AWS expanded their partnership with a dedicated low-latency interconnect between OCI and AWS for multicloud networking. The addition of Exascale confirms that both vendors have chosen to stop positioning themselves as direct competitors in the enterprise data market — and are instead jointly targeting organizations that refuse to choose between their legacy Oracle stack and their AWS ecosystem.
What CIOs and Data Teams Should Take Away
The implications go beyond pricing. This model shift changes the nature of architecture decisions: workloads that were previously offloaded to lower-performance databases for cost reasons can now be brought back onto Exadata without locking in fixed capacity. For organizations carrying large Oracle database estates and looking to modernize their data infrastructure on AWS, ExaDB-XS significantly reduces the financial risk of migration — and puts performance back at the center of the equation, without requiring a multi-year commitment to access it.