AWS–Azure Multicloud Interconnect Caps Preview at 1 Gbps, No Production SLA

AWS and Azure built a direct pipe between clouds — with sharp limits

Microsoft's Azure Multicloud Interconnect for AWS entered public preview this week, offering a managed, private Layer 3 connection between an Azure virtual network and an AWS VPC. The service avoids the public internet and uses MACsec encryption, but enterprises should not mistake preview access for production readiness. The connection is capped at 1 Gbps per customer, covers only four region pairs, carries no SLA, and has no published general-availability pricing from either vendor.

The 1-Gbps ceiling makes the service unsuitable for high-throughput production workloads such as cross-cloud database replication, disaster recovery with continuous sync, or large-scale AI training pipelines that move datasets between providers. AWS and Microsoft have stated an ambition to support up to 100 Gbps after general availability, but that is a future commitment, not current capacity. Enterprises running latency-sensitive or bandwidth-heavy cross-cloud architectures still need dedicated interconnection services or customer-built combinations of AWS Direct Connect and Azure ExpressRoute.

What this means for multicloud buyers

The preview period waives Azure service fees and data-egress charges on the Azure side, and AWS has also made its portion free during the trial. That creates a zero-cost window to test cross-cloud architectures without incurring the usual egress penalties that make multicloud designs expensive. Buyers can use the preview to validate whether keeping analytics, databases, or AI workloads split between AWS and Azure improves performance or cost compared to consolidating on one provider.

But procurement teams should model the eventual data-transfer price, not the free-preview economics. Neither AWS nor Microsoft has published final bandwidth-tier pricing or per-gigabyte transfer charges for general availability. Egress and inter-region networking fees routinely change the total cost of a multicloud design by 20% or more, so any architecture tested during the preview needs a cost model that includes realistic transfer pricing at production scale.

The service competes directly with dedicated interconnection providers including Equinix Fabric, Megaport, and PacketFabric. AWS and Microsoft are pulling part of that integration into their own control planes, which could weaken the role of independent connectivity providers for straightforward AWS–Azure links. Enterprises that already contract with third-party interconnect vendors should compare the hyperscaler-managed option against their existing agreements once pricing and SLAs are published, but the preview version does not yet offer enough capacity or reliability guarantees to replace production-grade dedicated connectivity.

Cloud infrastructure spending hit $143.4 billion in Q2 2026

Synergy Research Group measured worldwide cloud-infrastructure-services spending at $143.4 billion in Q2 2026, up 43% year over year. AWS held 28% market share, Microsoft Azure 20%, and Google Cloud 15%, representing 63% of global revenue combined. Google Cloud grew 82% and Oracle Cloud Infrastructure grew 121% in the quarter, both expanding faster than AWS and Azure from smaller bases.

The growth figures give large buyers stronger grounds to pursue multicloud bidding and capacity diversification rather than defaulting to a single provider. Oracle's 121% growth makes OCI more credible in evaluations involving Oracle databases, regulated workloads, or negotiated AI capacity, though the available data does not show whether that growth rate is sustainable or what workload types drove it.

The 43% market expansion also signals continuing pressure on enterprise infrastructure budgets, particularly for GPU-heavy AI deployments. Buyers should separate ordinary compute growth from AI-related consumption before using market growth rates in budget forecasts, because AI infrastructure carries materially different cost structures and capacity constraints than general-purpose compute.

Google Cloud scored highest in Forrester's current-offering evaluation

Forrester's Q3 2026 public-cloud evaluation named Google Cloud a Leader and awarded it the highest score in the current-offering category, with top marks in 23 of 30 criteria including AI development services, databases, analytics, containers and Kubernetes, modernization, and security. The result gives Google Cloud an externally validated positioning advantage in technical breadth, particularly for Kubernetes-heavy, data-intensive, or AI-platform procurements.

The Forrester evaluation is not a performance benchmark or customer-adoption measure. It should inform shortlisting rather than determine a purchase by itself. Enterprises still need to validate regional availability, committed-use discounts, support costs, and migration complexity before selecting a provider, but the report establishes that Google Cloud should be included in competitive evaluations rather than treated solely as a secondary hyperscaler.

What to watch

The AWS–Azure interconnect will remain unsuitable for production-critical workloads until Microsoft and AWS publish SLAs, expand regional coverage, raise bandwidth limits, and clarify pricing. Enterprises testing the preview should document current data-transfer volumes and map them to potential GA pricing scenarios before committing to a cross-cloud architecture that depends on this service.

The wider market trend is clear: buyers have more credible alternatives for negotiating price, availability, and AI capacity than they did 18 months ago. Google Cloud and Oracle Cloud Infrastructure are growing fast enough to function as competitive leverage in AWS and Azure deals, and the Forrester evaluation provides external validation for including Google Cloud in technical evaluations. Procurement teams should treat multicloud as a negotiating strategy, not just an architecture pattern.