Microsoft Azure customers from India to the UAE noticed sluggish performance and timeouts on September 6, 2025, as multiple fiber-optic cables in the Red Sea were severed, forcing internet traffic onto longer, often congested detours. The cable damage in this critical maritime corridor between Asia, Africa, and Europe disrupted the shortest data paths for many Azure services, particularly those crossing the Middle East. While Microsoft’s rapid traffic rerouting prevented a full-blown outage, the incident exposed persistent structural vulnerabilities in the global internet backbone that cloud-dependent enterprises can no longer ignore.

The physical internet relies on a fragile underbelly

The internet is as much a physical network as a logical one. Over 500 submarine fiber-optic cables, laid on the ocean floor, carry more than 95% of all international data, underpinning everything from cloud platforms to financial markets and streaming services. These cables are concentrated in a handful of narrow maritime corridors, with the Red Sea and the Bab al-Mandeb strait serving as a chokepoint connecting South and East Asia to Europe and the Middle East. Dozens of cables converge there, and when several are damaged simultaneously, the shortest paths disappear, forcing traffic onto alternative routes that can add hundreds of milliseconds of latency and strain already-loaded links.

What happened on September 6

At around 05:45 UTC on September 6, 2025, automated routing telemetry and internet monitoring groups such as NetBlocks detected abrupt path changes and performance degradation for traffic transiting the Red Sea. Carrier telemetry pointed to multiple cable faults near Jeddah, Saudi Arabia. Regional internet users in Pakistan, India, the UAE, and Saudi Arabia reported slower speeds, jitter on voice calls, and intermittent connectivity drops. Microsoft quickly posted an Azure Service Health advisory warning customers of “increased latency” for traffic that previously traversed the Middle East, clarifying that traffic not dependent on that corridor should be unaffected.

From cable cut to cloud incident—the technical anatomy

When a submarine cable breaks, a predictable chain reaction unfolds. First, the physical break removes primary capacity. The Border Gateway Protocol (BGP) then reconverges, and routers select alternate paths—often sending packets thousands of kilometers in the wrong direction. Propagation delay jumps, and if the alternate paths were not provisioned for the sudden volume, queuing delays, jitter, and packet loss follow. Latency-sensitive workloads—video calls, synchronous database replication, chatty APIs—degrade first. Microsoft and other cloud providers maintain separate management backbones, which is why Azure’s control plane remained reachable even while user-facing data flows suffered.

Microsoft’s measured response

Rather than declare a sweeping outage, Microsoft’s service advisory was carefully worded: “You may experience increased latency for traffic that previously traversed through the Middle East.” This aligns with industry practice of distinguishing between availability (services still up) and performance (slower than normal). Engineers began rerouting traffic and rebalancing capacity across remaining cables, but stressed that “undersea fibre cuts can take time to repair.” The advisory remained the primary communication channel, allowing customers to self-diagnose and apply mitigations.

The real-world toll: from end users to enterprise SLAs

End users in affected regions faced sluggish web browsing, buffering video streams, and rising complaint volumes. For enterprises running cross-region Azure workloads, the impact was more insidious: API latencies spiked, backup and replication windows stretched, and real-time services flirted with SLA breaches. A synchronous database replica between Southeast Asia and Europe, for instance, may have suddenly operated with three times the usual commit delay, risking application timeouts. While some large organizations had pre-negotiated traffic-engineering options, many were left scrambling.

Why repairs take weeks—and what’s being done

Repairing an undersea cable is a logistics-heavy maritime operation. Specialized cable-repair vessels must localize the fault, travel to the site—often in contested or busy waters—lift the damaged section, perform a mid-sea splice, and test the link. Permitting, security, and ship availability can stretch this from days to weeks. Initial reports pointed to candidate cables like SEA-ME-WE-4 and IMEWE, though consortiums had yet to publish formal fault reports at press time. Investigations will likely focus on anchor strikes from commercial shipping, accidental fishing damage, or—given regional tensions—deliberate sabotage. The International Cable Protection Committee notes that anchor strikes account for a large share of annual faults, but in politically sensitive waters, attribution remains tricky and repair timelines unpredictable.

Resilience shown, but structural risks persist

The incident highlighted the strength of automated routing and multi-path architectures: reachability was largely preserved, averting a catastrophic blackout. Yet three structural risks were laid bare. First, geographic concentration: multiple cables share the same Red Sea landing corridors, so a single event can take out several at once. Second, repair capacity: the global fleet of cable-repair ships is small, and regional security constraints delay access. Third, attribution and security: ambiguity about cause complicates both operational response and policy debates about protecting critical infrastructure.

Immediate steps for IT teams

For enterprises dependent on Azure or any cloud with traffic traversing the Red Sea, the following actions can limit damage:

  • Map exposure: identify which cross-region flows rely on the affected corridor.
  • Harden timeouts and retries: back off aggressive retry logic and increase timeouts for cross-region calls.
  • Defer large transfers: delay bulk data migrations, backups, or container image pulls until path performance stabilizes.
  • Validate failovers: test alternate-region failovers that avoid the chokepoint, ensuring data consistency.
  • Demand transparency: ask cloud providers and carriers for path-level visibility and negotiate contingency capacity in contracts.

Long-term architecture for a fragile world

The event underscores that virtual N+1 redundancy does not equal physical diversity. Architects must explicitly plan for alternate seafloor and land routes. Synchronous replication across continents should be replaced with asynchronous models where possible, and edge caching with distributed CDNs can reduce the need for long-distance round trips. In contracts, enterprises should press for rapid temporary transit leases during disruptions.

A policy wake-up call

Governments and international bodies are already responding. The International Telecommunication Union (ITU) has launched advisory initiatives to improve submarine cable resilience, including faster repair processes and multistakeholder protection regimes. The UK Parliament’s inquiry into undersea cable security explicitly flagged the risk of simultaneous damage to multiple links. The Red Sea incident will likely accelerate investment in new cable landings, route diversity, and diplomatic efforts to safeguard maritime chokepoints. For the cloud industry, it’s a reminder that geopolitics and physical infrastructure are inseparable from digital services.

Conclusion

The September 6 cable cuts that raised latency for Microsoft Azure users were not a platform outage, but they were a shot across the bow. Even the most ethereal cloud services rest on a web of glass threads stretched across ocean floors—threads that are vulnerable to anchors, antagonism, and accidents. Microsoft’s traffic engineering limited the pain, but could not eliminate it. For enterprises, the takeaway is stark: submarine cable risk is real operational risk. For policymakers and the cloud industry, the task is to accelerate diversity, repair capability, and protection before the next cable slip turns a performance hiccup into a connectivity crisis.