Oracle Database@AWS Gets Sub-Millisecond Networking and Reaches Twelve AWS Regions
Updated: 6 days ago
Title: Oracle Database@AWS Gets Sub-Millisecond Networking and Reaches Twelve AWS Regions
Date: 13 April 2026
Type: Blog
Author: SAASiQ (contact@saasiq.ai)
Word count: 1411 words
Reading time: 6 min
Published: 13-04-2026
AWS said on 1 April that applications running on EC2 can now reach Oracle Database@AWS with a network round trip consistently under a millisecond, at no extra charge, in six regions. On 8 April it said the service was available in twelve AWS regions, with London among the additions, and Oracle told investors in March that it expects 22 to be live by the end of May. Linking Oracle's own cloud to AWS is a separate matter, and for now it still runs through third-party exchanges rather than a direct Oracle interconnect of the kind Oracle has with Microsoft Azure and Google Cloud.
What the new networking does
AWS calls the feature high performance networking. When a customer creates an ODB network, the private network inside its AWS account where Oracle Database@AWS sits, AWS now attaches an EC2 placement group to it automatically. Instances launched into that placement group are put physically close to the Exadata infrastructure in the same Availability Zone, which reduces the number of network hops and, according to AWS's documentation, keeps the round trip from application to database consistently below a millisecond.
It covers Exadata VM clusters and Autonomous VM clusters on dedicated Exadata infrastructure, but not Autonomous Database Serverless. It applies only to ODB networks created on or after 1 April, so an existing deployment does not pick it up on its own. The placement group can be shared with other accounts in the same AWS organisation through AWS Resource Access Manager, and it works with the usual EC2 purchasing options (On-Demand Capacity Reservations, Savings Plans and Reserved Instances) and with ECS and EKS through launch templates.
The six launch regions were Ohio, Canada Central, Frankfurt, Dublin, Tokyo and Sydney, and AWS said more would follow. Standard EC2 charges still apply to the instances themselves; the optimised placement costs nothing extra.
Why the round trip matters
AWS aimed the feature at payment processing, securities trading and other high-volume transaction systems that customers run on Exadata on-premises today. Oracle Database@AWS was sold from the start as a way to move Exadata and Real Application Clusters systems to AWS like for like, and systems of that kind make large numbers of short database calls, each of which waits for a network round trip. The time spent on the network is multiplied by the number of calls in each transaction.
The placement group deals with where the application servers sit relative to the Exadata, which adding regions does not. AWS's documentation says it reduces both the number of network hops and the variation in latency, and presents it for workloads that need predictable, low-latency connectivity to Exadata databases, with manufacturing execution systems added to the list alongside payments and trading.
Twelve regions and a timetable
Oracle Database@AWS became generally available on 8 July 2025 in North Virginia and Oregon, with 20 more regions planned. Ohio, Frankfurt and Tokyo followed, then Canada Central and Sydney on 3 February, which brought the total to seven, and Dublin on 27 February. AWS's 8 April announcement took the count to twelve with London, Mumbai, Hyderabad and Seoul, although it also lists Dublin as new despite the February launch. Dublin, Mumbai and Hyderabad have two Availability Zones, London and Seoul have one, and Canada Central and Sydney have moved to two, which gives customers in those countries the option of a standby database in a separate zone without leaving the region.
On Oracle's third-quarter call on 10 March, co-chief executive Clay Magouyrk said Oracle began the quarter with two AWS regions live, ended it with eight and would end the fourth quarter, which closes on 31 May, with 22. He put the equivalent counts at 33 live regions with Microsoft and 14 with Google.
Oracle also renamed the three services on 24 March, to Oracle AI Database@AWS, @Azure and @Google Cloud. AWS's own announcements still use the old name.
How it is bought and run
Oracle runs the Exadata infrastructure, on OCI hardware deployed inside AWS data centres, and support comes jointly from the two companies. A customer requests a private offer from Oracle through AWS Marketplace, then creates and manages databases in the AWS console. At general availability AWS and Oracle said the spending counts toward a customer's existing AWS commitments, that existing Oracle licences can be brought across, and that Oracle Support Rewards apply.
For a buyer, the commitment point affects how the purchase is funded. A company that has already signed a large spending commitment with AWS can run Oracle's database there and draw down money it has committed anyway, rather than open a separate cloud contract with Oracle.
The service includes Exadata Database Service and Autonomous Database, with Real Application Clusters. AWS lists integrations with AWS Key Management Service, CloudWatch, IAM, CloudFormation, VPC Lattice and EventBridge, zero-ETL feeds into Amazon Redshift, and backups to Amazon S3. Fidelity Investments, Nationwide and SAS were named as early customers.
Connecting OCI itself to AWS
Oracle Database@AWS puts the database inside AWS. A customer that also runs workloads in OCI's own regions still needs a network path between the two clouds, and here Oracle's arrangements with AWS differ from those with Microsoft and Google.
Oracle and Microsoft announced a direct interconnect on 5 June 2019, starting between Oracle's Ashburn region and Azure US East; it pairs an OCI FastConnect circuit with an Azure ExpressRoute circuit. Oracle Interconnect for Google Cloud pairs FastConnect with Google's Partner Interconnect, and Oracle's documentation says neither company charges ingress or egress fees on it.
For AWS, Oracle's reference architecture links an OCI FastConnect circuit and an AWS Direct Connect circuit through a virtual router at Equinix, and Megaport sells an equivalent service. Traffic stays off the public internet, but the customer pays three parties: Equinix for the router and the virtual connections, AWS for Direct Connect and data egress, and Oracle for FastConnect. Oracle's guide recommends cities where all three are present and names London, Frankfurt, Ashburn and Tokyo.
AWS's own product for this kind of link, AWS Interconnect - multicloud, went into preview on 30 November 2025 in five AWS regions, with Google Cloud as the first partner and Microsoft Azure due later in 2026. The Register reported bandwidth of 1 Gbps during the preview, rising to 100 Gbps at general availability. AWS published the API on GitHub so that other cloud providers can adopt it, and its announcement did not name Oracle.
The database underneath
Oracle AI Database 26ai was announced at Oracle AI World on 14 October 2025 as the replacement for Database 23ai, delivered through the October 2025 release update rather than as an upgrade. Enterprise Edition for on-premises Linux x86-64 followed in the January 2026 release update, 23.26.1.
At its AI World Tour in London on 24 March, Oracle announced an MCP server for Autonomous AI Database, which lets outside agents and MCP clients query the database without custom integration code. It also announced Deep Data Security, which applies each end user's access rules inside the database, so an AI agent acting for a user sees only the rows, columns and cells that user is allowed to see.
Oracle's numbers and Japan
Oracle reported third-quarter revenue of $17.2 billion on 10 March, up 22 per cent. Cloud infrastructure revenue rose 84 per cent to $4.9 billion, and multicloud database revenue grew 531 per cent, for which Oracle gave no dollar figure. Remaining performance obligations, the contracted revenue not yet recognised, reached $553 billion. Oracle guided to capital spending of $50 billion for the fiscal year ending in May, on revenue of $67 billion.
In Japan, SoftBank said in October 2025 that it would launch Cloud PF Type A, its own cloud and AI service built on Oracle Alloy, in its eastern Japan data centre in April 2026 and in western Japan in October 2026. Alloy lets a partner run OCI in its own data centres and sell the services itself. SoftBank said its version would offer more than 200 OCI services, with data and systems managed and operated inside Japan.
What happens next
SAASiQ's view is that customers already running latency-sensitive systems on Oracle Database@AWS should check when their ODB network was created, because only networks created on or after 1 April get the placement group.
Oracle expects 22 AWS regions to be live by 31 May. AWS has said high performance networking will reach more regions and that Azure will join AWS Interconnect - multicloud later in 2026, without giving dates for either.
SAASiQ - Intelligent Solutions for SaaS ©
