AWS and Azure Launch Direct Private Interconnect, Bypassing Equinix and Megaport

AWS and Microsoft Enter the Interconnection Business

Amazon Web Services and Microsoft Azure moved a direct multicloud networking service into public preview in early September. The offering provides private, MACsec-encrypted Layer 3 connectivity between AWS and Azure regions, with a stated target of up to 100 Gbps per connection once generally available.

The service competes directly with Equinix, Megaport, Google Cloud, telecom carriers, and software-defined networking providers that currently dominate the cloud interconnection market. AWS and Microsoft are attempting to capture part of the connectivity layer themselves, reducing dependence on neutral exchange providers and keeping more of the multicloud stack under their control.

For enterprise buyers, the value proposition is straightforward: eliminate internet exposure for cross-cloud traffic, simplify network architecture, and improve latency for workloads running across both providers. The private connection replaces routing through the public internet or configuring multiple third-party peering points.

The strategic risk is equally clear. AWS and Microsoft now control another dependency point for enterprises running active workloads across both clouds. Buyers relying on this service for critical cross-cloud replication or data synchronization will need service-level commitments, failover plans, and visibility into per-connection pricing and cross-cloud data-transfer charges before treating it as a replacement for carrier or exchange-based connectivity.

The available reporting does not include a general availability date, public price list, or region-by-region rollout schedule. The 100 Gbps figure is a capacity claim, not an independently published performance benchmark. Enterprises should verify preview-region availability and encryption-management responsibilities before committing production traffic.

Multicloud Remains Hybrid, Not Portable

Flexera's 2026 State of the Cloud Report found that only 14% of surveyed organizations operate exclusively in a multicloud environment without private cloud infrastructure. The figure matters because it separates genuine public-cloud portability from the more common pattern of using AWS, Azure, and on-premises systems together.

This hybrid reality benefits vendors that manage public and private infrastructure from a single control plane: VMware by Broadcom, Red Hat, IBM, and Nutanix. It creates less demand for strategies built entirely on native hyperscaler tooling, which typically lack on-premises integration.

Buyers planning multicloud should budget for a control plane spanning public and private environments, including policy management, security, cost allocation, identity, and observability. The 14% figure suggests that "multicloud" typically means hybrid infrastructure management, not interchangeable workloads that move freely between AWS and Azure.

Enterprises should define whether their goal is provider negotiation leverage, disaster recovery, regulatory data placement, workload optimization, or actual portability. Each objective requires different architecture and produces different migration costs. A disaster-recovery failover to Azure does not require the same investment as maintaining active production workloads across AWS and Google Cloud with continuous data synchronization.

Operating Costs Remain a Material Penalty

Halkwinds Research reports that 89% of organizations with more than $5 million in annual cloud spending use at least two cloud providers in production, based on a survey of 524 enterprise cloud teams. Its 2026 cost benchmark says multicloud environments cost 23% more to operate than single-cloud environments while delivering 41% better availability.

The 23% operating premium comes from duplicated tooling, skills, networking, security controls, governance, and data movement. Enterprises running Kubernetes on AWS and Azure need separate monitoring stacks, separate security policies, separate cost-management tools, and engineers trained on both platforms. Data moving between clouds incurs egress charges that single-cloud architectures avoid.

The 41% availability improvement may justify the extra cost for revenue-critical systems. An e-commerce platform that loses $50,000 per minute of downtime can absorb a 23% operations premium in exchange for cross-cloud failover. An internal HR system with four-hour recovery-time objectives probably cannot.

Buyers should evaluate multicloud by workload class instead of applying it uniformly. High-availability systems may justify active-active or cross-cloud designs. Internal applications may be cheaper and simpler on one provider. The business case depends on downtime cost, compliance requirements, and negotiation leverage, not a blanket assumption that multicloud always reduces risk.

What This Means for Enterprise Buyers

Network architecture is becoming a competitive battleground. AWS and Azure offering direct private connectivity could pressure Equinix, Megaport, and carrier-based providers, but preview status and undisclosed pricing leave the commercial case incomplete. Enterprises should track general availability, pricing announcements, and region expansion before replacing existing interconnection contracts.

The cost of multicloud remains measurable and substantial. A 23% operations premium is not trivial. Buyers need per-workload business cases that account for duplicated tooling, skills, and data movement. Availability improvements justify the cost only when downtime has a quantifiable revenue impact.

Multicloud is not the same as portable. Most enterprises use multiple clouds because they also run private infrastructure, not because they move workloads freely between AWS and Azure. Portability requires investment in abstraction layers, common data formats, and cross-cloud orchestration. Using two clouds does not create portability by default.