FR
live

Google Cloud and Azure link their clouds in minutes through Cross-Cloud Interconnect

On November 18, 2026, Google Cloud extended its partner Cross-Cloud Interconnect to Microsoft Azure in preview, offering a private encrypted link provisioned in minutes across three regions. For a team running both GCP and Azure, physical interconnection is now a choice rather than an obligation.

A clean fiber-optic jumper cable bridging the gap between two rows of server cabinets.

November 18, 2026. Google Cloud extends its partner Cross-Cloud Interconnect to Microsoft Azure, in preview. November 18, 2026. The service is jointly supported by Google Cloud and Microsoft, and is provisioned in minutes from the console or the API. November 18, 2026. It opens in three regions: us-east4 (N. Virginia), us-west2 (Los Angeles), and europe-west3 (Frankfurt). Why it matters: connecting two clouds used to mean weeks of physical coordination; Google Cloud and Azure just turned that project into an API call.

What the service changes

The starting point is a well-known point of friction. Connecting two clouds meant ordering physical connections from each provider, configuring networking on both sides, and coordinating across multiple teams — a process that could take weeks, or even months. The new service removes that chain: you pick a bandwidth from the Google Cloud console or API, and both clouds handle the rest, with no hardware to order and no manual network configuration.

Traffic does not cross the public internet. It is encrypted automatically between the Google Cloud and Azure edge routers, with no customer-side encryption configuration. Reliability is built in: the connection runs across physically separate locations for redundancy, and both providers jointly monitor the link and handle maintenance and recovery.

Pricing breaks with the classic interconnection model. Google Cloud announces a fixed price for the connection’s bandwidth, which makes costs predictable — a contrast with bespoke quotes and cross-charges of physical builds. On the Microsoft side, the service is called Azure Multicloud Interconnect, and the Azure networking team describes it as a “private, managed path” between the two clouds.

Three regions, and a preview that says what it says

The preview is not a worldwide rollout. The service opens in three regions — us-east4, us-west2, and europe-west3 — alongside the already generally available offerings to AWS and OCI. Google Cloud promises to add regions as the Azure offering moves toward general availability.

This is the point to read soberly: for a production workload between GCP and Azure, you are today limited to North America and Frankfurt. Teams whose workloads live in Asia-Pacific or South America will have to wait. The “minutes, not weeks” promise is real, but it applies to a still-narrow regional footprint.

The pricing detail should be taken at face value too: Google Cloud communicates a fixed price for bandwidth, but the announcement publishes no price grid, no bandwidth tiers, and no full region list. Real cost figures will come from the Google Cloud quote — and from the Azure bill on the other side, which remains the under-covered half of the announcement.

One asymmetry is worth flagging: the fixed price covers the Google Cloud side of the link. The Azure Multicloud Interconnect bill sits on the Microsoft side, and the announcement publishes neither its price nor its bandwidth tiers, so budget for both clouds rather than the headline number alone.

The real news: multicloud friction disappearing

This is not one more networking feature. It is the removal of friction that has long made multicloud expensive to operate. Enterprises that spread their models, data, and applications across multiple clouds did so while accepting an integration tax; that tax just dropped to a few minutes of provisioning.

The quotes Google Cloud surfaces make the point in terms of time and cost. Walmart, through its engineering director Gerald Bothello, says it “shortened time to production and reduced overall costs” by simplifying connectivity between providers. Nestlé, through Ralf Huebenthal, points to a “reliable, scalable foundation” for connecting workloads across environments. These are CFO arguments as much as architect arguments: multicloud becomes an architecture decision, not a networking budget.

The service sits inside an expanding ecosystem. partner Cross-Cloud Interconnect is built on an open specification for network interoperability, published last year by Google Cloud in collaboration with AWS, which any provider can adopt. The AWS and OCI offerings are already generally available, and a last-mile variant with Lumen covers, in preview, customers connecting their private clouds, on-premises sites, and neocloud environments to Google Cloud.

What it changes for your architecture

Three use cases change cost in practice. First, disaster recovery: replicating data between GCP and Azure no longer requires a dedicated physical interconnect, but an on-demand provisioned link. Second, AI pipelines: teams that train on one cloud and serve on another gain a private, high-bandwidth path, where the public internet posed both latency and compliance problems. Third, environment mergers: inheriting a fleet on the other cloud no longer forces a choice between an expensive repatriation and a flimsy VPN tunnel.

The limit to keep in mind is the asymmetry of control. The link is jointly managed, but you pay two bills and depend on two SLAs. For a critical workload, a dedicated physical interconnect still makes sense when guaranteed throughput and direct support matter more than speed of setup. The preview targets the middle segment: teams that need a private link without wanting to manage hardware.

Beyond performance, the off-internet path carries a compliance benefit: traffic between regions stays inside the providers’ private backbone, which matters for data-residency requirements that forbid crossing the public internet. The joint monitoring also means both vendors own the link, so an incident has two accountable parties rather than a gray zone between two separate contracts.

The choice against the alternatives

Three options coexist for linking two clouds. A site-to-site VPN remains the most common: simple, but it crosses the public internet, its latency varies, and its bandwidth plateaus quickly on replication workloads. A dedicated physical interconnect offers guaranteed throughput and direct support, but takes weeks to build, costs real money, and requires a presence in an exchange point. Cross-Cloud Interconnect occupies the middle: a private, encrypted, redundant path provisioned in minutes, but billed at a fixed bandwidth price rather than contractually guaranteed throughput.

The consequence is a shift in the decision. The VPN becomes the choice for getting started and lightweight environments; the physical interconnect, the choice for critical workloads under hard SLA; and Cross-Cloud Interconnect, the choice for the in-between case — replication, AI pipelines, and production multicloud — which was precisely the most poorly served until now. The Azure preview closes the last major provider gap after AWS and OCI.

The verdict

If you run workloads between GCP and Azure in the served regions — North America or Frankfurt —, provision a Cross-Cloud Interconnect link rather than building a physical interconnect: the setup time and the price predictability argue for the preview. The use case that justifies switching today is data replication and multi-cloud AI pipelines, where a private path changes both latency and compliance.

If your workloads live outside the three regions, or you require contractually guaranteed throughput with dedicated support, stay on a physical interconnect — the preview covers neither your geography nor your SLA requirements. The takeaway is the direction: Google Cloud and Azure have lowered the cost of multicloud entry to the point where the network stops being the argument against spreading workloads.

References

The cyber brief, every Tuesday

The flaws that matter and the patches to apply, in a ten-minute read.

No spam. One-click unsubscribe.
read next

On the same topic

AWS Lambda adds GPU instances and brings inference into serverless

On November 18, 2026, AWS let you run Lambda functions on EC2 G-series instances, with no driver, framework, or scaling management, using ready-made vLLM, SGLang, and PyTorch images. For intermittent inference and batch media processing, serverless GPU is finally workable.

← Back to the feed

Type at least two characters.

↑ ↓ navigate ↵ open esc dismiss