Amazon Web Services this month opened a private, direct network connection between its cloud and Oracle Cloud Infrastructure, the second cloud provider to join a service called AWS Interconnect - multicloud since it reached general availability in April 2026. Microsoft Azure, the third provider AWS named as a future partner, has not yet joined. Together the three moves show hyperscalers building direct plumbing between each other's networks, something they spent the last decade routing customers away from.
What AWS Interconnect actually does
AWS Interconnect - multicloud is a managed Layer 3 private connection between a customer's AWS VPC and a resource on another cloud provider's network. AWS reached general availability with Google Cloud as its first partner in April 2026, publishing the details in its own weekly product roundup and a same-day post on the AWS News Blog. Each connection AWS provisions comes with four redundant physical paths across two locations, BGP routing configured automatically, and MACsec encryption and jumbo frames turned on by default. Bandwidth runs from 1 Gbps to 100 Gbps and can be resized from the console without re-provisioning the link. AWS backs the Google Cloud connections with a 99.99% availability SLA, according to AWS's own documentation.
The pricing model is the detail worth pausing on. AWS charges a flat hourly rate scaled to the bandwidth tier the customer selects, with no per-gigabyte data transfer charge on the interconnect itself, per AWS's published multicloud pricing page. Most complaints about multicloud architecture center on egress fees eating any savings from running workloads on a second provider. A flat, predictable hourly charge is a different economic model, and it is the first time AWS has offered one for cross-cloud traffic. Starting in May 2026, each account also gets one free 500 Mbps interconnect per cloud provider per region, per AWS's what's-new announcement.
Google Cloud's side of the same link went generally available in the same window, described in a Google Cloud blog post as an extension of its existing Cross-Cloud Interconnect product to work with the AWS-built connection-coordinator protocol. Five region pairs are live: AWS's Virginia, Northern California, Oregon, London and Frankfurt regions each connect to a matching Google Cloud region.
Oracle's turn, and why it matters to database-heavy shops
Oracle and AWS announced on April 16, 2026 that they would extend the same interconnect model to Oracle Cloud Infrastructure, framed around Oracle's existing Oracle AI Database@AWS offering, which already lets customers run Oracle Database workloads inside AWS data centers. That April announcement described the connectivity as coming "later in 2026." AWS moved faster than that language suggested: by mid-May, AWS's what's-new page listed the OCI connection as available in preview in the US East (N. Virginia) region, the same region where the Oracle-AWS database service already runs.
For any enterprise running Oracle Database workloads alongside AWS compute, a large and specific population given how many finance, ERP and supply-chain systems still run on Oracle, this closes a real gap. Previously, connecting an Oracle-hosted database to an application tier on AWS meant either a third-party network provider like Megaport or Equinix, a VPN over the public internet, or accepting the latency and egress cost of routing through Oracle's own or AWS's own edge. A managed, redundant, encrypted link removes that middle step. It also means Oracle no longer needs AWS to build a special integration for a single customer relationship; the connectivity is a standing product both sides can point new customers to.
The open specification, and why Azure hasn't shown up
AWS published the protocol behind Interconnect - multicloud as an OpenAPI 3.0 specification on GitHub under an Apache 2.0 license, under the name aws/Interconnect. Any cloud provider can implement the same connection-coordinator handshake AWS already runs with Google Cloud and is now running in preview with Oracle. That is a deliberate move to make the protocol look neutral rather than AWS-owned, even though AWS designed it and hosts the reference implementation.
Azure is conspicuously absent. AWS named Microsoft as a future partner when it announced general availability in April, but as of the mid-May Oracle preview, no Azure connection has shipped, no preview date has been set, and Microsoft has not confirmed a timeline publicly. Microsoft runs its own competing connectivity product, Azure ExpressRoute, and has less incentive to make it easy for customers to route around Azure entirely. Whether Microsoft adopts AWS's open specification, builds a competing standard, or simply stays out, is now the open question in this story.
What to watch next
The next concrete marker is whether AWS's OCI connection moves from preview to general availability before the end of 2026, matching the roughly four-month gap between the Google Cloud preview and its GA. A second marker: whether Microsoft responds to the published specification at all, either by adopting it or by announcing a competing interconnect of its own. Platform teams running genuinely multicloud estates, particularly ones with Oracle Database workloads, have a real reason to test the OCI preview now rather than wait, since preview availability is currently limited to a single region.
Sources: AWS Interconnect GA · AWS Interconnect OCI preview · Oracle-AWS multicloud networking · InfoQ · Google Cloud blog · AWS Interconnect pricing · aws/Interconnect on GitHub