Microsoft Azure customers across the Middle East and Asia faced significant latency spikes and service degradation on September 6, 2025, after multiple undersea fiber-optic cables in the Red Sea were severed simultaneously. The incident forced cloud traffic onto longer, congested detours, exposing the brittle physical underpinnings of the internet and the limits of cloud resilience when the seabed itself becomes a single point of failure.

What Happened

Early on September 6, automated monitoring systems and carrier bulletins reported faults on at least two major submarine cable systems in the Red Sea corridor near Jeddah, Saudi Arabia. The SMW4 (South East Asia–Middle East–Western Europe 4) and IMEWE (India–Middle East–Western Europe) cables, which together carry a substantial portion of traffic between Asia, the Middle East, and Europe, went dark. Subsequent telemetry confirmed that other nearby cables may also have been damaged, though operators have not publicly identified every system.

Microsoft posted an Azure Service Health advisory the same day, warning that “customers who connect to their services from or to the Middle East may experience increased latency.” The company said its engineers were rerouting traffic, rebalancing capacity, and closely monitoring the situation. While most services remained reachable, the performance hit was immediate and measurable: round-trip times (RTTs) jumped by tens to hundreds of milliseconds depending on the alternate path taken.

The Technical Anatomy: Why a Seabed Break Surfaces as a Cloud Problem

The path from a broken cable on the seafloor to a user-visible slowdown is short and deterministic. When a subsea segment is severed, Border Gateway Protocol (BGP) updates propagate across carrier networks, and routers reconverge to alternate next-hops. Traffic that once followed the shortest, highest-capacity trunk now takes longer routes—sometimes looping around Africa, traversing congested terrestrial backhaul, or relying on lower-capacity satellite links.

The extra distance increases propagation delay, while additional hops add processing and queuing delays. In the Red Sea incident, rerouted traffic likely traveled via the Cape of Good Hope, through terrestrial routes across the Levant, or via Pacific cables to the Americas and back to Europe. Each alternative adds 50–200 ms of latency, enough to break time-sensitive applications. Packet loss and jitter also climbed as overloaded backup links dropped or queued packets.

This incident is a textbook demonstration of how cloud performance is bound to physical layer constraints. Azure’s software-defined backbone and global edge network can steer traffic programmatically, but they cannot conjure new fiber in real time. The laws of physics—particularly the speed of light in glass—define the minimum possible delay. Microsoft’s own latency dashboard and independent probes from ThousandEyes and RIPE Atlas confirmed the degradation lasting for hours until engineering mitigations stabilized alternate paths.

Microsoft’s Response and Mitigation

Microsoft’s immediate response followed an established playbook. Network engineers adjusted BGP policies to deprecate routes transiting the damaged segments and to prefer alternate subsea systems and terrestrial cross-connects. Capacity teams shifted workloads to underutilized links, leased additional transit from third-party carriers, and selectively prioritized control-plane and latency-sensitive customer traffic. Edge caches and regional Points of Presence (PoPs) also helped absorb some demand.

The company committed to daily Azure Service Health updates—or sooner if conditions changed—while repair planning and ship mobilization progressed. By the following day, Microsoft reported that most services had returned to normal as the rerouted paths stabilized and interim capacity arrangements took hold. However, the advisory remained active because full restoration of physical fiber capacity depends on maritime repair operations, which can take days to weeks. Cable faults in the Red Sea are notoriously difficult to repair quickly due to security risks, shipping traffic, and limited availability of specialized repair vessels.

Regions and Services Affected

The latency spike primarily affected cross-region flows that would normally transit the Red Sea corridor: Asia–Europe, Asia–Middle East, and intra-Middle East traffic. Customers in India, Pakistan, the UAE, Kuwait, and Saudi Arabia reported the most visible impact, as documented by network monitoring firms and social media reports. Azure’s own telemetry showed elevated latency for services hosted in UAE North, Qatar Central, and India Central regions when communicating with counterparts in Europe.

Latency-sensitive workloads bore the brunt: VoIP calls stuttered, video conferences froze, online gaming sessions lagged, and synchronous database replication jobs either slowed to a crawl or timed out entirely. Bulk data transfers and asynchronous backups were delayed but generally completed. The control plane—the APIs for managing resources—remained responsive, and regionally contained services (those without cross-region dependencies) saw little or no impact.

The Verified Timeline

Based on carrier updates, Microsoft’s own advisories, and independent network telemetry, the key timestamps are:

  • September 6, early morning UTC: Automated systems detect route flaps and sudden capacity drops in the Red Sea corridor.
  • September 6, 10:00 UTC: Microsoft publishes its first Azure Service Health incident under tracking ID LMVT-2K8, warning of elevated latency and ongoing rerouting.
  • September 6–7: Carriers and Microsoft implement BGP changes and capacity rebalancing; some customer-facing services see gradual improvement.
  • September 8: Microsoft issues an update stating that “the majority of services have returned to normal levels of latency” but that monitoring continues.
  • Ongoing: Physical repair operations are expected to take weeks; customers should remain vigilant for possible residual latency.

What Remains Uncertain

Despite detailed reporting, several key facts remain unverified or in dispute. Outlets like Reuters and Livemint cited figures suggesting that “up to 17% of global internet traffic was disrupted,” but these numbers are often calculated using aggregate bandwidth estimates and may not reflect actual user impact. Without raw data from cable operators or neutral monitors, such percentages should be treated with caution.

The root cause is also unconfirmed. Early speculation pointed to accidental damage from a ship’s anchor, but no official investigation has concluded. Geopolitical tensions in the Red Sea, including Houthi attacks on commercial vessels, have raised the specter of deliberate sabotage, though there is no public evidence. Until the cable owners release forensic findings, any attribution is premature.

Guidance for Azure Customers

For companies relying on Azure, the incident offers immediate and long-term lessons. Short-term actions include:

  • Check Azure Service Health: Microsoft’s advisory is the authoritative source for customer-specific impact.
  • Validate exposure: Identify applications that use cross-region paths involving the Middle East or South Asia. Review replication and backup schedules that traverse these routes.
  • Harden timeouts and retries: Increase conservative retry backoffs for cross-region API calls; reduce chatty synchronous interactions.
  • Defer non-urgent transfers: Postpone bulk cross-region data movements until routes stabilize.
  • Engage support: For business-critical flows, discuss temporary alternatives with Microsoft, such as moving edge endpoints or triggering multi-region failovers that avoid the affected corridor.

Long-Term Lessons for Cloud Architects

The Red Sea incident reinforces several structural truths about cloud resilience that are too often ignored:

  • Logical redundancy isn’t physical diversity. Multiple virtual network paths may share the same physical chokepoint. Architects must map real-world transit geometry—knowing which subsea systems and landing stations their traffic traverses.
  • Latency budgets matter. Applications designed for perfect network conditions fail under even moderate latency increases. Resilience testing must include network stress scenarios: simulate 100–200 ms of added RTT and 0.5–2% packet loss.
  • Demand transparency from providers. Ask cloud and carrier partners for documentation on physical route diversity. Contractual guarantees of “no shared chokepoint” for critical circuits are essential.
  • Invest in application-level patterns. Circuit breakers, exponential backoff, idempotent APIs, and regional caches reduce the blast radius of network disruptions.

Industry and Geopolitical Implications

Subsea cables are the internet’s spine, and the Red Sea is one of its narrowest, most fragile vertebrae. Over 90% of intercontinental data traffic travels through fewer than 500 active cable systems, many concentrated in a handful of geological and geopolitical choke points—the Luzon Strait, the Strait of Malacca, the Suez Canal region, and the Red Sea. The latter is especially perilous because of intense maritime traffic, armed conflict, and limited repair vessel availability.

Repair operations can be delayed by insurance hurdles, port access permissions, and the sheer time needed to locate, retrieve, and splice deep-sea fiber. The global fleet of cable repair ships is small—fewer than 60 vessels—and often booked months in advance. In a security-conscious environment like the Red Sea, military escorts may be required, adding another layer of complexity.

The incident will likely accelerate calls for investment in more diverse subsea routes. Projects like Google’s Blue-Raman cable, which bypasses the Red Sea by crossing the Mediterranean and then traversing Saudi Arabia overland, exemplify a trend toward hybrid architectures. Regulators and industry consortia are also exploring faster damage-detection systems and “cable corridors” that physically separate new cables from shipping lanes.

The Bigger Picture: Cloud Is Just Someone Else’s Computer—On Someone Else’s Cable

The Azure latency event is a stark reminder that the cloud’s abstraction of infinite, resilient compute does not extend to the physical transport layer. Every cloud provider, no matter how large, is ultimately a tenant on shared subsea cables and terrestrial fiber networks that are vulnerable to anchors, earthquakes, sabotage, and geopolitical strife.

Microsoft’s response—rapid rerouting, transparent communication, and capacity rebalancing—was textbook, and it kept services from failing outright. But the uncomfortable truth is that the only long-term fix—physical route diversity—requires years of planning and billions of dollars. In the interim, enterprises must architect their workloads for an internet where the next cable cut is not a matter of if, but when.

Conclusion

The September 2025 Red Sea cable cuts caused measurable pain for Azure customers but, crucially, did not escalate to a platform-wide outage. The incident exposed the deep dependency of even the most sophisticated cloud platforms on a brittle, centuries-old maritime chokepoint. For Microsoft, it validated its investment in traffic engineering and global capacity. For the industry, it underscored the urgent need to diversify subsea routes, accelerate repair capacities, and embrace true physical redundancy—not just logical abstractions. For IT leaders, the takeaway is clear: test your applications against the real-world physics of a hemispherical detour, because someday soon, they may have to take one.