The Root Cause: A Direct Hit on Utility Power

The outage was triggered by a single, powerful external event: a severe thunderstorm. Unlike some cloud failures that stem from internal software bugs or configuration errors, this was a physical infrastructure problem from the start. Microsoft's impact statement was clear: the storm caused "widespread utility power loss across multiple West US 2 datacenter facilities." This led to what the company initially described as a "full-region impact," as it was unable to immediately confirm the failure was isolated to specific availability zones.

A Wide-Ranging Service Meltdown

The loss of power did not just affect one or two services. Microsoft reported a broad, region-wide degradation that impacted a significant portion of its portfolio. The official status history listed the following affected services :

• AI and Machine Learning: Azure OpenAI Service
• Core Compute: Azure Virtual Machines, Azure Kubernetes Service (AKS), Azure Functions
• Databases and Analytics: Azure SQL Database, Azure Cosmos DB, Azure Database for PostgreSQL / MySQL flexible servers, Azure Data Factory, Azure Databricks
• Management and Identity: Azure Resource Manager, Azure Policy
• Storage and Backup: Azure Storage Accounts, Azure NetApp Files, Azure Backup
• Integration and Messaging: Azure Service Bus, Azure Communication Services
• Other Key Services: Log Analytics, Azure App Service, Azure Maps, Azure Migrate

The Fog of War: Backup Power Systems Did Not Fail

A critical point of clarification is that, unlike some infamous datacenter outages, there is no evidence from the available sources that the backup power systems failed in this event. Microsoft's status updates explicitly stated that "datacenter power has been fully restored," indicating that the initial utility loss was the problem, not a subsequent failure of on-site generators or Uninterruptible Power Supplies (UPS). The service degradation was a result of the initial power cut and the subsequent time required to bring complex storage and network infrastructure back online safely and reliably.

This stands in contrast to other documented Azure incidents. For example, during a West US outage in February 2026, a failure of a voltage transformer caused a rapid voltage rise that propagated downstream and damaged UPS units, leading to a complete loss of power. No such internal hardware failure has been reported for the May 29 event.

The Arc of Recovery

The recovery from a region-wide power loss is a marathon, not a sprint. Based on the official Azure status updates, here is the timeline of key events:

• 04:27 UTC: Outage begins. A severe thunderstorm causes widespread utility power loss across multiple West US 2 datacenters.
• Recovery Progress: Datacenter power is fully restored. Following this, the majority of network and storage infrastructure is brought back online. Infrastructure in Availability Zone 2 recovers and operates normally.
• Lingering Impact: The recovery efforts become concentrated on a subset of infrastructure within Availability Zones 1 and 3, where two storage stamps require final recovery and validation.
• Final Known State ("Active – Recovering"): As of the last available status snapshot, the incident was still marked as "Active – Recovering," meaning a full resolution had not yet been declared for all affected services.

A Broader Pattern of Weather-Related Disruptions

This latest outage is far from an isolated incident. It is part of a troubling pattern of weather-related vulnerabilities that have plagued Microsoft’s Azure platform for years. These incidents demonstrate that even the most sophisticated cloud infrastructure is ultimately at the mercy of physical reality.

• September 2018, South Central US: A severe weather event near San Antonio, Texas, including lightning strikes, caused a voltage sag that knocked cooling systems offline in a datacenter. Servers overheated and shut down, causing a multi-day outage for some services due to architectural dependencies.
• February 2026, West US: A localized power interruption at a datacenter was caused by a failure in an on-site substation transformer. While backup systems engaged, the initial voltage surge damaged UPS equipment, leading to a more complex recovery process that impacted the Microsoft Store and Windows Update services.

These precedents share a common thread with the May 29 event: an external physical event directly impacts the datacenter's power chain, leading to a cascading failure of the digital services that depend on it.

The Stakes for AI Reliability

The inclusion of the Azure OpenAI Service in the list of degraded services marks a notable escalation in the impact of these outages. As enterprises increasingly embed AI models into critical workflows and customer-facing products, regional cloud downtime is no longer just an infrastructure inconvenience. It directly disrupts AI-powered applications, search, and automation. While a severe thunderstorm is an act of nature, the resulting outage highlights a growing concentration risk for AI workloads reliant on specific cloud regions.

Microsoft had not yet published a detailed Root Cause Analysis (RCA) for this specific incident within the provided source window. As such, the final conclusions on the precise sequencing of power restoration and why certain storage stamps required extended recovery time will be confirmed only with an official post-incident review.

The Root Cause: A Direct Hit on Utility Power

The outage was triggered by a single, powerful external event: a severe thunderstorm. Unlike some cloud failures that stem from internal software bugs or configuration errors, this was a physical infrastructure problem from the start. Microsoft's impact statement was clear: the storm caused "widespread utility power loss across multiple West US 2 datacenter facilities." This led to what the company initially described as a "full-region impact," as it was unable to immediately confirm the failure was isolated to specific availability zones.

A Wide-Ranging Service Meltdown

The loss of power did not just affect one or two services. Microsoft reported a broad, region-wide degradation that impacted a significant portion of its portfolio. The official status history listed the following affected services :

• AI and Machine Learning: Azure OpenAI Service
• Core Compute: Azure Virtual Machines, Azure Kubernetes Service (AKS), Azure Functions
• Databases and Analytics: Azure SQL Database, Azure Cosmos DB, Azure Database for PostgreSQL / MySQL flexible servers, Azure Data Factory, Azure Databricks
• Management and Identity: Azure Resource Manager, Azure Policy
• Storage and Backup: Azure Storage Accounts, Azure NetApp Files, Azure Backup
• Integration and Messaging: Azure Service Bus, Azure Communication Services
• Other Key Services: Log Analytics, Azure App Service, Azure Maps, Azure Migrate

The Fog of War: Backup Power Systems Did Not Fail

A critical point of clarification is that, unlike some infamous datacenter outages, there is no evidence from the available sources that the backup power systems failed in this event. Microsoft's status updates explicitly stated that "datacenter power has been fully restored," indicating that the initial utility loss was the problem, not a subsequent failure of on-site generators or Uninterruptible Power Supplies (UPS). The service degradation was a result of the initial power cut and the subsequent time required to bring complex storage and network infrastructure back online safely and reliably.

This stands in contrast to other documented Azure incidents. For example, during a West US outage in February 2026, a failure of a voltage transformer caused a rapid voltage rise that propagated downstream and damaged UPS units, leading to a complete loss of power. No such internal hardware failure has been reported for the May 29 event.

The Arc of Recovery

The recovery from a region-wide power loss is a marathon, not a sprint. Based on the official Azure status updates, here is the timeline of key events:

• 04:27 UTC: Outage begins. A severe thunderstorm causes widespread utility power loss across multiple West US 2 datacenters.
• Recovery Progress: Datacenter power is fully restored. Following this, the majority of network and storage infrastructure is brought back online. Infrastructure in Availability Zone 2 recovers and operates normally.
• Lingering Impact: The recovery efforts become concentrated on a subset of infrastructure within Availability Zones 1 and 3, where two storage stamps require final recovery and validation.
• Final Known State ("Active – Recovering"): As of the last available status snapshot, the incident was still marked as "Active – Recovering," meaning a full resolution had not yet been declared for all affected services.

A Broader Pattern of Weather-Related Disruptions

This latest outage is far from an isolated incident. It is part of a troubling pattern of weather-related vulnerabilities that have plagued Microsoft’s Azure platform for years. These incidents demonstrate that even the most sophisticated cloud infrastructure is ultimately at the mercy of physical reality.

• September 2018, South Central US: A severe weather event near San Antonio, Texas, including lightning strikes, caused a voltage sag that knocked cooling systems offline in a datacenter. Servers overheated and shut down, causing a multi-day outage for some services due to architectural dependencies.
• February 2026, West US: A localized power interruption at a datacenter was caused by a failure in an on-site substation transformer. While backup systems engaged, the initial voltage surge damaged UPS equipment, leading to a more complex recovery process that impacted the Microsoft Store and Windows Update services.

These precedents share a common thread with the May 29 event: an external physical event directly impacts the datacenter's power chain, leading to a cascading failure of the digital services that depend on it.

The Stakes for AI Reliability

The inclusion of the Azure OpenAI Service in the list of degraded services marks a notable escalation in the impact of these outages. As enterprises increasingly embed AI models into critical workflows and customer-facing products, regional cloud downtime is no longer just an infrastructure inconvenience. It directly disrupts AI-powered applications, search, and automation. While a severe thunderstorm is an act of nature, the resulting outage highlights a growing concentration risk for AI workloads reliant on specific cloud regions.

Microsoft had not yet published a detailed Root Cause Analysis (RCA) for this specific incident within the provided source window. As such, the final conclusions on the precise sequencing of power restoration and why certain storage stamps required extended recovery time will be confirmed only with an official post-incident review.