AWS and Google Cloud Just Made Multicloud Real: What the Interconnect Deal Changes

Two Rivals Build a Bridge

For a decade, the big three cloud providers competed on lock-in. Proprietary networking, bespoke APIs, and punishing egress fees kept workloads sticky. Enterprises that wanted to run services across AWS and Google Cloud had to thread traffic through third-party network fabrics, negotiate separate colocation agreements, and wait weeks for circuit provisioning.

That changed in late 2025 when AWS unveiled AWS Interconnect — multicloud at re:Invent, with Google Cloud as the launch partner. Google simultaneously extended its Cross-Cloud Interconnect to support the same capability from its side. The result: a managed, private, high-speed network link between an Amazon VPC and a Google Cloud VPC that either party can provision in minutes through their own console.

The service entered public preview across five region pairs — US East (N. Virginia), US West (N. California), US West (Oregon), Europe (London), and Europe (Frankfurt) — each paired with a corresponding Google Cloud region. During preview, customers can create a 1 Gbps connection per account at no cost, with bandwidth expected to scale up to 100 Gbps at general availability.

What makes this more than a networking feature is the open specification. AWS and Google published the Connection Coordinator API as an OpenAPI 3.0 spec in a public GitHub repository, explicitly inviting other providers and partners to adopt it. Microsoft Azure is expected to join in the second half of 2026, turning a bilateral deal into an emerging industry standard.

Why Now — and Why It Matters

The timing is not accidental. Three forces pushed the hyperscalers toward collaboration.

Regulatory pressure. The EU Data Act, which took full effect in late 2025, requires cloud providers to eliminate switching barriers and support interoperability. Google had already waived egress fees for customers migrating away, and AWS followed. Building a native interconnect is the next logical step — it signals compliance and preempts further regulatory action.

Enterprise demand. According to the Flexera State of the Cloud 2026 report, 89% of enterprise organizations now run a multicloud strategy, with vendor lock-in prevention consistently ranking among the top motivations — cited as the primary driver by 42% of respondents, and as a contributing factor by up to 68% in broader surveys. Yet most multicloud deployments are accidental — different teams choosing different providers — not architected. A managed interconnect gives enterprises a way to run intentional multicloud without a networking PhD.

AI workload distribution. As companies deploy inference pipelines, retrieval-augmented generation systems, and fine-tuning jobs, they increasingly need to place workloads where the right GPU inventory, pricing, or specialized services exist. A private cross-cloud link makes it feasible to split a pipeline across providers without routing sensitive data over the public internet.

How It Works Under the Hood

The architecture is deliberately simple. From the AWS side, a customer creates an Interconnect multicloud connection specifying the target provider (Google Cloud), the destination region, and the desired bandwidth. AWS provisions a dedicated circuit over its backbone to a point of presence shared with Google Cloud. Google’s Cross-Cloud Interconnect completes the link on its side. Traffic never touches the public internet.

Both sides integrate with existing networking constructs. On AWS, the connection plugs into Transit Gateway or Cloud WAN. On Google Cloud, it attaches to a Cloud Router via a VLAN attachment. Customers manage routing, security groups, and firewall rules exactly as they would for any VPC peering — no new abstractions to learn.

The open Connection Coordinator API handles the handshake between providers. It defines how one cloud requests a connection, how the peer acknowledges it, how bandwidth is allocated, and how the link is torn down. Because the spec is public, third-party network-as-a-service providers like Megaport or Equinix could implement it to offer alternative transport paths.

What This Means for Vendor Lock-In

It would be naive to declare vendor lock-in dead. The interconnect solves the network layer — it does not port your DynamoDB tables to Bigtable or translate Lambda functions into Cloud Functions. Proprietary managed services remain the deepest source of lock-in, and neither AWS nor Google has any incentive to commoditize those.

What changes is the cost of optionality. Before, even evaluating a second cloud required months of networking setup. Now, an enterprise can spin up a cross-cloud link in minutes during preview and test whether splitting a workload across providers delivers real gains. The friction of experimentation drops to near zero.

This matters for several strategic scenarios:

• Best-of-breed AI stacks. A company might train models on Google Cloud’s TPU pods and serve inference on AWS Inferentia — and the cross-cloud link makes the data pipeline between them private and fast.
• Compliance-driven data residency. When a specific region or provider has the required compliance certifications, the interconnect lets enterprises place data where regulations demand while keeping the application layer on their primary cloud.
• Disaster recovery without duplication. Active-passive DR across two clouds becomes architecturally cleaner when the providers offer a managed private link rather than a VPN over the public internet.

The Pricing Question Nobody Can Answer Yet

The most important detail is still missing: general-availability pricing. During preview, the 1 Gbps connection is free. But enterprise multicloud architectures will need 10–100 Gbps of sustained bandwidth, and the per-GB data transfer charges on both sides will determine whether this is genuinely transformative or just a well-marketed science project.

For context, cross-cloud data transfer today costs between $0.01 and $0.09 per GB depending on the provider and region. Organizations consistently report that data integration and transfer costs represent a significant share of their total cloud spend, often rivaling primary compute and storage costs. If the managed interconnect carries a premium over existing third-party fabric options, large enterprises may stick with Megaport, Equinix, or PacketFabric for the cost advantage.

The egress fee question is also unresolved. Google waived egress fees for switching under EU Data Act pressure, but operational data transfer fees remain. Whether interconnect traffic receives discounted egress pricing — and how deeply — will shape adoption more than any technical feature.

Azure Joins Next — Then What?

Microsoft Azure is expected to integrate with AWS Interconnect in the second half of 2026, which would complete the big-three triangle. Once all three hyperscalers support the same open Connection Coordinator API, the networking layer of multicloud becomes a solved problem — at least in theory.

The real test will be whether the open spec attracts participation beyond the big three. Oracle Cloud Infrastructure, Alibaba Cloud, and regional providers could adopt it to plug into the same fabric. If they do, the spec becomes a true interoperability standard. If they don’t, it remains a bilateral convenience between the dominant players.

For enterprise architects, the strategic calculus is shifting. The question is no longer whether multicloud is technically feasible but which workloads justify the operational overhead of running across providers. With a managed interconnect, the answer set just got considerably larger.

Frequently Asked Questions

Sources & Further Reading

• AWS and Google Cloud collaborate on multicloud networking — Google Cloud Blog
• AWS and Google Cloud Preview Secure Multicloud Networking — InfoQ
• AWS finally moves to simplify multicloud operations with Google — Network World
• Build resilient and scalable multicloud connectivity architectures with AWS Interconnect — AWS Blog
• Azure follows Google in joining AWS multicloud initiative — Intercept Cloud
• AWS, Google link up to ease multicloud deployments — CIO Dive