Multiple undersea fiber-optic cables serving the critical Red Sea corridor suffered simultaneous cuts early Saturday, September 6, 2025, triggering a continent-scale reroute of Microsoft Azure traffic that has left Asia-to-Europe cloud connections battling elevated latency and jitter. Microsoft acknowledged the incident at 05:45 UTC with a Service Health advisory confirming that while Azure compute and control-plane services remained fully operational, customers traversing the affected path “may experience higher-than-normal latency.”

The disruption affected at least five major submarine cable systems—SMW4, IMEWE, EIG, AAE-1, and SEACOM/TGN-EA—according to monitoring data and carrier advisories, yanking the shortest fiber routes between the Arabian Peninsula and the Suez Canal out of service. For Azure, every packet that would have wound through those systems had to be hurried onto the next-best alternative paths, often thousands of kilometers longer and already shared with displaced traffic from other networks. Microsoft’s engineering teams immediately rerouted flows, but the sheer physics of longer paths and congested alternate links turned a physical infrastructure event into a persistent, if partial, cloud performance degradation.

The corridor at the center of the internet

The Red Sea and its approaches to the Suez Canal form one of the world’s most concentrated chokepoints for intercontinental data traffic. More than a dozen high-capacity submarine cable systems snake through these narrow waters, carrying the majority of digital communications between Europe, the Middle East, and Asia. The global internet relies on about 1.4 million kilometers of such cables, but their geographic concentration in a few corridors—the Red Sea, the Malacca Strait, the trans-Pacific routes—creates systemic risk. When several cables in the same corridor fail simultaneously, the remaining paths become instant bottlenecks, shoving latency upward even if no packets are lost.

Saturday’s cuts removed key systems like the Southeast Asia-Middle East-Western Europe 4 (SMW4) and the India-Middle East-Western Europe (IMEWE) cable, along with the Europe India Gateway (EIG), Asia Africa Europe-1 (AAE-1), and SEACOM/TGN-EA. Together these cables handle terabits per second of commercial, scientific, and consumer traffic. Their severed strands meant that cloud providers like Microsoft had to herd data onto detours that might loop through the Pacific, swing across terrestrial networks in Europe, or squeeze through alternative submarine corridors with far less headroom.

What Microsoft told customers

At 05:45 UTC on September 6, Microsoft posted a Service Health advisory. The company said its network performance had begun to degrade due to external cable cuts and that customers moving traffic between Asia and Europe through the Middle East “may experience increased latency.” Engineering teams had already started dynamic rerouting and were “continuously monitoring, rebalancing, and optimizing routing” to limit impact. The advisory characterized the event as a performance degradation, not a platform-wide outage, and promised daily updates—or sooner if conditions changed.

Behind the scenes, Azure’s software-defined networking stack kicked in. Border Gateway Protocol (BGP) reconvergence redirected prefixes away from the dead segments, and traffic engineering tools spread the load across the remaining links to avoid overloading any single path. These automated steps are why service stayed up. Yet Microsoft was candid that the mitigations could only do so much: “Users are still facing delays and slower-than-usual connections,” a company statement noted. The operational playbook—reroute, rebalance, communicate—proved effective at stopping a hard outage, but it could not repeal the laws of physics.

Regional ripple effects

The cable failures were not a Microsoft-only event. National carriers across South Asia and the Middle East reported degraded throughput. In the United Arab Emirates, telecoms Du and Etisalat (e&) acknowledged that home broadband and mobile services were “running more slowly” on Saturday evening. Internet monitoring group NetBlocks confirmed that connectivity in Pakistan and India had also dipped, with many websites and apps failing to load. Pakistan Telecommunications issued its own advisory that capacity on the SMW4 and IMEWE systems was reduced. Independent route monitors logged spikes in latency and route changes consistent with multiple co-located fiber breaks near Jeddah, Saudi Arabia.

For enterprises, the regional impacts translated directly into slower API responses, extended database replication windows, and poor VoIP or video call quality when endpoints spanned the Asia-Europe divide. Even applications that had not been designed for low-latency operation began to feel the pinch as round-trip times ballooned by tens to hundreds of milliseconds. The cable cuts effectively turned a high-speed express route into a long, congested backroad—and every packet felt the delay.

Why latency, not outage, is the story

A full-blown outage would have cut off entire regions from Azure. Here, the cloud remained reachable; the problem was that data took much longer to get from A to B. That distinction matters because many enterprise architectures are not engineered for extended latency variation. Synchronous replication between datacenters, for example, often assumes an upper bound on round-trip time. When latency crosses that threshold, replication can fail, transaction logging can stall, and high-availability clusters can flap. Similarly, chatty microservices that depend on dozens of quick API calls across a continent may slow to a crawl as each call adds 100 ms of extra propagation delay.

BGP reconvergence introduces its own transient instability. When multiple routes are withdrawn or shifted, the internet’s routing tables can flap, creating short bursts of packet loss and blackholed traffic until stable paths settle. Cloud providers like Microsoft tune their BGP timers and traffic engineering to minimize these transients, but end-user applications and their retry logic can amplify the visible effects. The result is a service that looks “up” on a status dashboard but feels painfully slow to the humans—or machines—relying on it.

The physical reality: repair takes weeks

Fixing a severed submarine cable is a major maritime operation. Specialized repair ships must sail to the fault location—often hundreds of kilometers from a port—locate the break using optical time-domain reflectometry, grapple the cable ends to the surface, splice in a new section, and then rebury it. The global fleet of cable-repair vessels is small, and their schedules are booked months in advance. Geopolitical tensions in the Red Sea region complicate matters: securing permits and ensuring crew safety can add days or weeks to the timeline.

Marine experts point out that 70–80% of subsea cable faults stem from accidental human activities such as dropped anchors or bottom trawling. Natural events like seabed shifts cause another 10–20%. The Red Sea has seen its share of suspected sabotage—Yemen’s Houthi rebels were accused in 2024 of planning attacks on undersea infrastructure—but attributing this specific incident remains premature. What is certain is that physical repairs will not be measured in hours or days. Microsoft and other cloud providers must therefore rely on software rerouting for weeks, not as a stopgap but as the primary defense.

Practical steps for Azure customers

IT teams whose applications cross the Asia-Europe corridor can take immediate, high-return actions while the routing instability persists. These measures focus on resilience and user experience:

  • Check Azure Service Health for subscription-level alerts. Microsoft’s public advisories contain the most current picture of mitigation efforts and expected recovery timelines. Set up Service Health alerts if you have not already.
  • Identify exposed workloads by mapping which services send data through the Middle East. High-risk candidates include backup replication, SQL Always-On groups in synchronous mode, and any cross-region VNet peering that transits the affected path.
  • Harden client and server timeouts with exponential backoff. Latency spikes can cause cascading failures if applications retry too aggressively. Ensure that SDKs and libraries are configured to handle extended delays gracefully.
  • Defer non-urgent cross-corridor transfers. Bulk data movements, cold backups, and development-environment syncs can wait until connectivity normalizes. This frees capacity for business-critical flows.
  • Use edge and CDN configurations to localize traffic. Azure Front Door, AWS CloudFront (for multi-cloud setups), or generic CDN providers can terminate user connections close to the user and cache content, reducing the need for long-haul round-trips.
  • Engage Microsoft support and your transit providers to explore temporary alternative connectivity, such as ExpressRoute circuits that might bypass the congested submarine corridors through terrestrial networks.

These steps are not silver bullets; they cannot restore the original physical routes. But they can shrink the blast radius and keep end users productive while the cables lie broken on the seafloor.

Beyond this incident: structural cloud resilience

The Red Sea cable cuts expose a structural vulnerability that goes beyond any single cloud provider. Logical redundancy—multiple availability zones, region pairs, and failover capabilities—can be undermined when those zone and regions all depend on the same physical cable corridor. A region pair in Europe and India may look independent on a topology diagram, but if the only fiber route between them is through the Red Sea, the “diverse” paths are an illusion.

Enterprise architects must start treating physical-path diversity as a first-class requirement. That means contracting with multiple subsea cable consortiums, using satellite links for critical control-plane traffic, and deploying edge compute nodes that keep latency-sensitive logic close to users. Regular failover testing should simulate not just a region going dark but also a prolonged period of high latency and packet loss on intercontinental links.

Cloud providers, for their part, are aware of the risk. Microsoft has invested in cross-continental paths that avoid the Middle East—for example, via the Pacific and the Atlantic—and its rerouting playbook shows that those investments pay off. Still, the incident is a stark reminder that the cloud’s “always-on” promise is ultimately anchored to ships, splices, and seabed cables.

Attribution: premature to point fingers

Within hours of the cable cuts, social media and some news outlets began speculating about deliberate sabotage. The Houthi movement in Yemen had previously threatened undersea infrastructure, and 2024 saw allegations—denied by the group—that rebels had attacked cables in the Red Sea. On this occasion, however, no credible forensic evidence has been made public. Microsoft and other operators limited their initial statements to operational impacts and repair plans. The International Cable Protection Committee notes that accidental damage from anchors and fishing gear is far more common than intentional attacks. Until cable consortiums publish post-mortem findings, attribution should be treated with healthy skepticism.

What enterprise IT leaders should do now

This incident is not a one-off anomaly. Subsea cable corridors will continue to concentrate traffic, repair ships will remain scarce, and geopolitical flashpoints will persist. Forward-looking organizations should:

  • Conduct a physical-route audit to map how their cloud traffic actually traverses the globe. Tools like traceroute, BGP looking glasses, and carrier route analytics can reveal hidden dependencies.
  • Build “corridor failure” playbooks that define escalation procedures, alternative routing options, and communication templates for stakeholders when a major subsea route breaks.
  • Reevaluate replication and synchronization modes for cross-region databases and storage. Asynchronous replication introduces data-lag but survives high-latency events; synchronous replication does not.
  • Shift latency-critical user-facing services to an edge-first architecture. By serving content and authentication from local CDN edges or distributed edge-compute platforms, you reduce the user-perceived impact of backhaul congestion dramatically.

Microsoft’s engineering teams performed as expected: the cloud stayed up, workloads kept running, and the damage was limited to degraded performance. The lesson for enterprises is that degraded performance can be just as damaging to business operations as an outright outage. The Red Sea cuts are a clear signal to harden against the physical realities that no amount of software abstraction can fully eliminate.