Microsoft and Amazon Web Services spent years selling customers on the idea that one cloud was enough. On September 1, 2026, both companies admitted that idea never really worked. They jointly announced Azure Multicloud Interconnect for AWS, a managed private network link that lets traffic move between Azure and AWS without touching the public internet. The service entered public preview on August 26, 2026, covering four region pairs, capped at 1 Gbps per customer, and shipping with no service-level agreement. It sounds modest. The numbers behind why it exists are not.
What Azure Multicloud Interconnect Actually Does
Strip away the branding and the service is a private pipe. Azure Multicloud Interconnect combines Microsoft’s existing ExpressRoute infrastructure with AWS’s Direct Connect backbone, then wraps both in a single managed resource that customers provision through one console instead of two. Traffic that used to cross the public internet, or require a customer to hand-build a circuit through a third-party colocation facility, now travels over a dedicated link that Microsoft and AWS operate jointly. A workload running in an Azure virtual network can reach a VPC in AWS US East without ever routing through a public gateway.
Microsoft’s Azure Networking team described the goal plainly in its launch post: “A jointly delivered, fully managed private connection between your Azure and AWS environments — without the burden of building and operating cross-cloud networking yourself” is how the company frames the pitch. That framing matters because it tells you who this is for: enterprises already running production workloads on both clouds, not startups picking a single provider.
Inside the September Announcement: Timeline and Technical Details
Microsoft’s engineering blog carries a publish date of August 26, 2026, which effectively started the preview quietly before the joint public rollout on September 1. AWS’s own Interconnect team echoed the announcement the same week, framing it as an extension of a specification AWS had already been building toward other providers. In its post, AWS wrote: “Microsoft and AWS are together introducing Azure Multicloud Interconnect, a jointly engineered, fully managed service that delivers private, high-throughput connectivity between Azure and AWS through a single logical resource”, describing a shared OpenAPI specification that both clouds’ provisioning tools can call.
The technical backbone is not new hardware. Microsoft’s own description makes that explicit: “Azure Multicloud Interconnect is a provider-managed, private intercloud connectivity service built on the proven foundations of Azure ExpressRoute and AWS Direct Connect”. In other words, Microsoft and AWS took two products that already existed separately and wired them together behind a shared control plane. Customers no longer need to provision an ExpressRoute circuit on one side and a Direct Connect port on the other, then manually pair them through a third-party exchange.
The Four Region Pairs Available in Preview
Preview access is narrow by design. Microsoft and AWS picked four region pairs that concentrate the bulk of dual-cloud enterprise traffic in North America, Europe, and Asia-Pacific, leaving large swaths of the map, including South America, most of Asia outside Sydney, and the Middle East, unaddressed for now.
| Azure Region | AWS Region | Bandwidth Cap (Preview) | SLA | Preview Cost |
|---|---|---|---|---|
| East US | US East (N. Virginia) | 1 Gbps | None | Fees waived |
| West US | US West (N. California) | 1 Gbps | None | Fees waived |
| Australia East | Asia Pacific (Sydney) | 1 Gbps | None | Fees waived |
| West Europe | Europe (Frankfurt) | 1 Gbps | None | Fees waived |
Microsoft and AWS have talked about a general-availability target of up to 100 Gbps per connection with 99.99% availability, but neither company has committed to a GA date. Until that ships, this is a test environment, not a production migration tool, and enterprises evaluating it should treat the 1 Gbps ceiling as a hard constraint for any serious data transfer workload.
Why Microsoft and AWS Are Doing This Now
Ten years ago this partnership would have been unthinkable. Microsoft and AWS have spent more than a decade fighting for the same enterprise IT budgets, and a joint engineering effort that makes it easier to leave either platform runs against the instinct to lock customers in. What changed is what workloads look like now. AI training and inference pipelines routinely pull data from wherever it lives, whether that’s a data lake on one cloud or a model endpoint on another, and enterprises stopped waiting for vendors to solve that problem cleanly.
AWS’s own networking team put the customer pressure bluntly: “Customers told us they wanted a better way to connect workloads spanning AWS and Azure, and the old ways of doing it were clunky”. That’s a notable admission from a company that has historically preferred customers stay inside its own walls. The explanation both companies give publicly centers on AI and data-heavy workloads, where moving large datasets between providers for training, inference, or backup has become routine rather than exceptional.
The Multicloud Numbers Behind the Decision
The business case rests on survey data that both vendors have cited internally for years. Flexera’s 2025 State of the Cloud Report found that 86% of organizations now run a multicloud environment, while 70% describe their strategy as hybrid, meaning they mix at least one public cloud with a private cloud or on-premises infrastructure. Only 30% run purely on public or purely on private infrastructure.
That 86% figure explains why Microsoft built a joint product with a direct rival instead of simply improving ExpressRoute on its own. If the overwhelming majority of large customers already split workloads across providers, the networking gap between those providers becomes a shared cost that both vendors absorb in the form of complexity and support tickets. Flexera’s broader cloud-spend tracking also shows AWS and Azure running close among large enterprises, with AWS holding a clearer lead among small and mid-sized businesses at 53% adoption versus 29% for Azure. That SMB gap gives Microsoft an added incentive: an easier bridge to AWS lowers the switching friction that keeps smaller customers away from Azure entirely.
How It Compares to AWS Direct Connect and Azure ExpressRoute
Before this launch, connecting Azure to AWS privately meant provisioning two separate products and stitching them together yourself. AWS Direct Connect gives you a dedicated link from your own network or a colocation facility into AWS. Azure ExpressRoute does the equivalent for Azure. Neither product was built to talk to the other directly, so enterprises needed a third party, typically a colocation provider, to physically cross-connect the two circuits in a shared facility.
Azure Multicloud Interconnect removes that middle step for customers who only need Azure-to-AWS connectivity. It is not a replacement for either underlying product. Both ExpressRoute and Direct Connect remain the foundation, and organizations that already rely on them for connections to other destinations, such as their own data centers, still need those circuits separately. What’s new is that Microsoft and AWS now offer a jointly provisioned path between just the two of them, cutting out the manual circuit-pairing work and the vendor coordination that used to take weeks.
MACsec Encryption and the 99.99% Availability Target
Both companies say the GA version will support MACsec link-layer encryption alongside the 100 Gbps bandwidth ceiling and 99.99% availability target. AWS framed the ambition directly: “With AWS Interconnect-multicloud and Azure Multicloud Interconnect, we’re proving what’s possible when both sides commit to a high bar: MACsec security out of the box, four-nines availability, and scalability at the click of a button”. None of that exists in the current preview, which carries no SLA and caps bandwidth at a fraction of what’s promised for GA.
No SLA During Preview — What That Means in Practice
Running production traffic over a link with zero uptime guarantee is a real risk, not a formality. Enterprises testing the preview should treat it strictly as a proof-of-concept environment: validate routing, test failover behavior, and measure latency, but keep the production path on existing ExpressRoute or Direct Connect circuits until Microsoft and AWS publish GA terms. The fee waiver during preview softens that risk financially, since neither the Azure service charge nor the Azure-side egress fee applies yet, but a free service with no SLA is still a free service with no SLA.
Where Google Cloud and Oracle Already Stand
Microsoft and AWS are not the first to ship this idea. Google Cloud’s Cross-Cloud Interconnect has offered private, managed connectivity between Google Cloud and external providers, including AWS and Azure, for longer than this new service has existed. Google built its version through its own infrastructure and partner network rather than a joint engineering deal with the other cloud, so the two approaches aren’t identical, but the end goal for customers is the same: fewer manual circuits, one bill, one support path.
Oracle Cloud Infrastructure moved even faster on the AWS side specifically. According to reporting from Kernel Talks, Oracle reached general availability on AWS’s interconnect specification about a month before Microsoft entered preview at the end of August 2026, making Azure the third major cloud to adopt that spec after AWS’s own infrastructure and Oracle. That sequencing matters for how this story gets told: Microsoft and AWS are not pioneering cross-cloud interconnects, they’re catching up to a pattern Oracle and Google had already validated with customers.
Megaport, Equinix, and the Neutral Fabric Squeeze
Before any cloud provider offered native cross-cloud interconnects, the job belonged to neutral network exchanges. Megaport and Equinix Fabric built businesses on letting enterprises provision software-defined cross-connects between any combination of clouds, data centers, and carriers from a single portal. Those platforms still offer something the new cloud-native services can’t match yet: true multi-cloud reach across dozens of providers rather than just two, plus a geographic footprint that spans markets none of the big three clouds have paired directly.
| Service | Provider(s) | Connection Model | Max Bandwidth (Target) | Key Limitation |
|---|---|---|---|---|
| Azure Multicloud Interconnect | Microsoft + AWS | Jointly managed, single resource | 100 Gbps (GA target) | Preview capped at 1 Gbps, no SLA, 4 regions |
| Google Cross-Cloud Interconnect | Google Cloud | Google-managed to external clouds | Up to 100 Gbps | Provisioned through Google’s side only |
| AWS Direct Connect | AWS | Dedicated link to AWS only | Up to 100 Gbps | Needs a separate circuit for the other cloud |
| Azure ExpressRoute | Microsoft | Dedicated link to Azure only | Up to 100 Gbps | Needs a separate circuit for the other cloud |
| Megaport / Equinix Fabric | Neutral network operators | Software-defined cross-connect | Up to 100 Gbps | Requires a third-party contract and port fees |
The squeeze on neutral fabrics is real but not immediate. Azure Multicloud Interconnect only covers Azure-to-AWS traffic in four regions during preview. Any enterprise that needs Azure connected to Google Cloud, Oracle, or a private data center in a market the new service doesn’t cover still needs Megaport or Equinix. The cloud-native option wins on simplicity for the single most common pairing. The neutral fabrics still win on breadth.
The AI Workload Angle: Why Cross-Cloud Bandwidth Matters Now
Every public justification Microsoft and AWS have given for this launch points back to AI. Training pipelines increasingly pull raw data from wherever it was collected, which for large enterprises is rarely a single cloud. A retailer might store transaction logs in AWS S3 while running its analytics and model-serving stack on Azure, or vice versa, because of a legacy contract, an acquisition, or simple negotiating leverage with each vendor. Moving terabytes of training data between those environments over the public internet is slow, unpredictable, and expensive at scale.
A private, high-throughput link changes that math, assuming the GA bandwidth targets hold. At a theoretical 100 Gbps, a large nightly data sync that might take hours over a constrained internet path could complete in minutes. That’s the pitch, and it’s consistent with a broader pattern this year of cloud vendors investing in dedicated infrastructure for AI data movement rather than treating it as ordinary traffic.
The diagram below illustrates the basic traffic path before and after this kind of interconnect exists, without referencing any specific vendor configuration syntax:
Before (manual circuit pairing):
Azure VNet -> ExpressRoute circuit -> Colocation cross-connect -> Direct Connect port -> AWS VPC
(separate contracts, separate billing, manual peering configuration)
After (Azure Multicloud Interconnect, preview):
Azure VNet -> Azure Multicloud Interconnect (single resource) -> AWS VPC
(shared provisioning API, one logical connection, 1 Gbps cap in preview)Pricing: What’s Free Today and What Isn’t Confirmed for GA
During preview, Microsoft is waiving the Azure service fee for the interconnect along with Azure-side data-egress charges, and AWS has described its side of the preview connectivity as free as well. Neither company has published final generally available pricing, a bandwidth-tier rate card, or confirmed per-gigabyte transfer costs. Reports describing a flat hourly bandwidth model with zero per-gigabyte charges have circulated but are not independently confirmed by either Microsoft or AWS, so enterprises should not budget around that assumption yet.
That pricing silence is worth watching closely. A common complaint about existing cross-cloud connectivity is that egress fees, not the connection itself, drive most of the real cost. If Microsoft and AWS ship GA pricing that still charges standard egress rates on top of a connection fee, the simplicity gain from this service won’t translate into meaningful savings for high-volume data movement. If they waive or discount egress specifically for interconnect traffic, that would be the more disruptive move, and neither company has said which direction they’re heading.
Historical Context: A Decade of Cloud Interconnect Attempts
Cloud providers have tried to solve cross-cloud networking before, usually by improving their own single-cloud product rather than partnering with a rival. AWS Direct Connect launched in 2011. Azure ExpressRoute followed a few years later. Both were built to connect customer infrastructure to one cloud, not to each other, because at the time most enterprises were still trying to pick a primary cloud rather than run several simultaneously.
From Direct Connect to Cross-Cloud Interconnect
The shift started when Google Cloud, as a distant third player in market share, had the most incentive to make interoperability with AWS and Azure easy, since forcing an all-or-nothing choice mostly benefited the two leaders. Google’s Cross-Cloud Interconnect reflected that competitive position. Oracle followed a similar logic as a smaller player chasing enterprise database workloads that often sit alongside AWS or Azure compute. Microsoft and AWS, as the two largest providers, had the least incentive to make switching easier, which is exactly why this joint announcement stands out. When the two biggest competitors in a market jointly build a bridge between themselves, it’s usually because customers made the old friction too expensive to defend.
Market Impact: What This Means for Network Vendors and Enterprises
For enterprise network architects, the immediate impact is limited by geography and bandwidth. Four region pairs at 1 Gbps won’t replace existing production links for companies moving serious data volumes today. The medium-term impact is about procurement leverage. Once GA pricing and SLAs land, enterprise network teams will have a genuine choice between a cloud-native joint service and a neutral fabric provider for the single most common cross-cloud pairing, and that choice will shape renewal negotiations with Megaport, Equinix, and regional colocation providers over the next renewal cycle.
For Microsoft and AWS themselves, the calculation is less about this specific product and more about defending share against Google Cloud and Oracle, both of which already had a cross-cloud interconnect story to tell enterprise customers evaluating multicloud strategies. Shipping a joint Azure-AWS option closes that gap rather than opening new ground. It also signals that both vendors now treat “avoid lock-in” as a selling point worth building for, rather than a risk to suppress.
Predictions: Where Multicloud Networking Goes From Here
- Azure Multicloud Interconnect will likely expand past its four initial region pairs within two to three quarters, prioritizing additional US and European pairings before broader Asia-Pacific coverage, following the same rollout pattern Microsoft has used for other Azure networking preview features.
- General availability will almost certainly arrive with published SLA terms closer to the stated 99.99% target, since neither Microsoft nor AWS typically leaves a jointly branded enterprise product in an SLA-free state for more than a few quarters.
- Egress pricing, not the connection fee, will be the detail that determines whether enterprises actually shift volume onto this service, and expect scrutiny on that point the moment GA pricing is published.
- Google Cloud and Oracle will likely respond with expanded region coverage or pricing adjustments to their own cross-cloud interconnect offerings rather than a direct feature-for-feature match, since both already have a head start on the underlying capability.
- Expect at least one more major cloud pairing announcement, most plausibly Google Cloud and AWS building a similarly tight joint product, within the next year, as competitive pressure makes standalone interconnects look increasingly dated next to jointly managed ones.
What This Doesn’t Solve Yet
It’s easy to read a joint Microsoft-AWS announcement as a finished product. It isn’t. The preview excludes most of the world’s data center regions, offers a fraction of the promised bandwidth, and carries no uptime guarantee. Enterprises running latency-sensitive or compliance-bound workloads across clouds still need to plan around those gaps for at least the next few quarters. The interesting part of this story isn’t the current state of the product, it’s the fact that it exists at all between two companies that spent a decade competing for the same workloads.
Frequently Asked Questions
What is Azure Multicloud Interconnect?
It’s a jointly managed private network service from Microsoft and AWS that connects Azure virtual networks to AWS VPCs without routing traffic over the public internet. It combines Azure ExpressRoute and AWS Direct Connect infrastructure behind a single provisioning resource.
When did Azure Multicloud Interconnect launch?
Microsoft published its technical announcement on August 26, 2026, and Microsoft and AWS jointly described the service publicly on September 1, 2026. It remains in public preview as of late September 2026.
Which regions support the preview?
Four region pairs: East US to US East (N. Virginia), West US to US West (N. California), Australia East to Asia Pacific (Sydney), and West Europe to Europe (Frankfurt).
How much does it cost?
During preview, Microsoft is waiving the Azure service fee and Azure-side egress charges, and AWS has described its side as free as well. Final general-availability pricing has not been published by either company.
Is there a bandwidth limit?
Preview connections are capped at 1 Gbps per customer per region pair. Microsoft and AWS have stated a general-availability target of up to 100 Gbps, but that has not shipped yet.
Does the preview come with a service-level agreement?
No. The preview ships without an SLA. Microsoft and AWS have discussed a 99.99% availability target for general availability, alongside MACsec encryption support, but neither is confirmed for the current preview period.
How is this different from AWS Direct Connect or Azure ExpressRoute?
Direct Connect and ExpressRoute each connect customer infrastructure to a single cloud. Azure Multicloud Interconnect uses both as its backbone but provisions the Azure-to-AWS link as one jointly managed resource, removing the need to manually pair separate circuits through a third-party colocation facility.
Does Google Cloud or Oracle offer something similar?
Yes. Google Cloud’s Cross-Cloud Interconnect has offered managed private connectivity to AWS and Azure for longer than this new service has existed, and Oracle Cloud Infrastructure reached general availability on AWS’s interconnect specification roughly a month before Azure entered preview.
