An immediate network response, a deferred budget unknown
An interconnection between Azure and AWS offered in preview illustrates a classic multicloud pattern: the technical building block addresses a network problem identified today, while its pricing trajectory remains open. Production traffic between two hyperscalers often takes less direct paths; a dedicated interconnection changes the equation for latency, predictability and path control. But the announcement concerns a preview, not a general service with a service-level commitment.
The question is therefore not only whether it works, but how much it will cost once it is no longer free. Distinguishing the two avoids building an architecture decision on a price that has not yet been published.
What Microsoft confirms, and what remains open
Microsoft confirms a preview at 1 Gbit/s, with no service fees or Azure egress fees for this interconnect during the preview, and no SLA. Three elements should be noted: the announced throughput, the scope of the free period, limited to the preview period and to this interconnect, and the absence of an availability commitment.
Traffic entering Azure requires an ExpressRoute gateway. This point has direct consequences: the gateway is not a configuration detail; it involves capacity and redundancy choices and associated costs that must be verified.
In Europe, Germany West Central is listed, not West Europe. The choice of region is therefore not neutral: it affects latency, data residency and the existing architecture.
Two unknowns remain: post-preview pricing and AWS-side costs. As long as they are not documented, no multi-year projection can be considered reliable.
Four cost lines to model
Before production traffic between Azure and AWS, I would ask for a cost model structured in four lines.
Capacity and ports. The 1 Gbit/s throughput announced for the preview is not necessarily the target throughput in production; sizing must be based on actual volume and include a margin.
Outbound data from each cloud. The announced free period covers Azure egress for this interconnect during the preview. It says nothing about AWS-side egress, which must be costed separately, nor about the regime applicable after the preview.
Ancillary network services. ExpressRoute gateway, any associated connectivity services, redundancy and monitoring fall under this line.
Operations. Engineering time, testing, incident management, documentation and monitoring of service conditions. This line is often forgotten even though it weighs on total cost of ownership.
Do not extend a promotional free period by default
A possible promotional free period must never be extended by default in the forecast. The reasoning is simple: a preview without an SLA and without published pricing offers no guarantee of pricing continuity. Extending the free period amounts to assuming that current conditions will persist, which has not been demonstrated.
The prudent method is to build at least two scenarios: one using the preview conditions, and one with an unknown post-preview price, handled as a sensitivity analysis. The break-even point is then identified: from what volume or what price the interconnection ceases to be profitable compared with alternatives.
As a hypothetical illustration, if the AWS-side egress cost exceeds the gain in latency or operational simplicity, the trade-off may tip. This is not a conclusion, only a criterion to test.
Decision criteria and questions to ask
What network problem does the interconnection solve, and is there a less costly alternative? Latency, throughput, compliance, path control: the need must be named before comparing options.
What outbound volume in each direction? The asymmetry of pricing regimes between Azure and AWS is the main uncertainty factor.
Is the Germany West Central region compatible with residency and latency requirements? Since West Europe is not listed, this may require an architecture review.
What is the full cost of the ExpressRoute gateway and its operation?
What happens if the preview ends, if conditions change or if the service does not reach production? An exit path must be planned.
The central question remains: does your interconnection decision remain profitable if a free period ends?
In summary
The Azure–AWS interconnection in preview is assessed on two levels: the immediate technical response and the budget trajectory. Microsoft confirms 1 Gbit/s, no service fees and no Azure egress fees during the preview, and no SLA; traffic entering Azure goes through an ExpressRoute gateway; Germany West Central is listed in Europe, not West Europe.
Post-preview pricing and AWS-side costs remain to be verified. As long as they are not, prudence consists in not integrating the free period as a default assumption and in explicitly documenting the scenarios retained.
The post behind this insight
Expanded from the LinkedIn post. The links below come from the original post; listing them does not imply independent verification.
LinkedIn ↗