Claude Goes Down Globally as AWS Data Centers Burn in the Middle East

Claude went down hard today. Starting around 11:49 UTC on March 2, Anthropic's flagship AI assistant began throwing errors across claude.ai, the developer console, Claude Code, and authentication services. The core API stayed operational, but everything that faces end users - the web interface, mobile app, and login system - went dark.

What Went Down

Anthropic's status page flagged the incident as "Elevated errors on claude.ai, console, and claude code" at 11:49 UTC. By 12:21 UTC, the company confirmed they were still investigating.

The breakdown by service, based on Downdetector reports:

Users hitting claude.ai saw one of two messages: "Claude will return soon. Claude is currently experiencing a temporary service disruption" or a blunt HTTP 529 error. Login and logout paths were specifically broken, meaning even users with active sessions couldn't re-authenticate.

Anthropic clarified one important distinction: "We have identified that the Claude API is working as intended. The issues we are seeing are related to Claude.ai and with the login/logout paths." That means developers calling Claude through the API - including production applications, Claude Code integrations via API keys, and enterprise deployments - were largely unaffected. The outage hit consumer and prosumer users hardest.

The AWS Middle East Situation

The timing is remarkable. Starting March 1, AWS experienced severe outages in its Middle East regions after unidentified objects struck a data center in the UAE (ME-CENTRAL-1), sparking a fire that forced emergency power shutdowns. The Bahrain region (ME-SOUTH-1) followed with its own power and connectivity failures hours later.

The numbers are significant: 38 AWS services went down in the UAE region and 46 in Bahrain, including EC2, Lambda, EKS, VPC, RDS, and CloudFormation. AWS warned that full recovery would "be many hours away."

The strikes on the data center coincided with Iranian retaliatory strikes on the UAE and Bahrain following US and Israeli strikes that killed senior Iranian officials. AWS hasn't confirmed or denied a connection between the geopolitical events and the facility damage.

Is the Claude outage related to the AWS Middle East failures? Anthropic hasn't said. Claude's primary infrastructure runs on AWS (Anthropic has a $4 billion partnership with Amazon), but the company's production workloads are understood to run mostly on US regions, not Middle East availability zones. The overlap in timing may be coincidental - or it may indicate cascading effects from AWS rerouting traffic or shifting capacity across regions.

The Reliability Question

This is not Claude's first outage in 2026. PiunikaWeb tracked four significant disruptions this year:

• January 14: Elevated error rates affecting Opus 4.5 and Sonnet 4.5
• January 22: Claude Code authentication issues
• February 4: HTTP 529 errors, failed fix attempt noted
• March 2: Today's outage - the most widespread so far

The pattern matters because Claude is increasingly embedded in production workflows. Claude Code now accounts for 4% of all GitHub commits. Enterprise teams use it for code review, document drafting, and data analysis. When Claude goes down, it doesn't just mean a chatbot is unavailable - it means CI/CD pipelines stall, developer workflows break, and business processes that have been rebuilt around AI assistance suddenly have a single point of failure.

Bloomberg reported the outage, noting thousands of affected users. Social media responses ranged from frustrated developers sharing error screenshots to the usual gallows humor about briefly reclaiming their lives from an AI assistant.

What This Means

Two things can be true simultaneously: Claude is one of the most capable AI assistants available, and its reliability isn't yet at the level that mission-critical infrastructure demands.

The fact that the API stayed up while the web interface and authentication went down suggests the failure was in Anthropic's application layer, not in the model serving infrastructure. That is a different kind of problem - and arguably more fixable - than a fundamental capacity or model issue. But for the millions of users paying $20 to $200 per month for Claude, "the API works fine" is cold comfort when the only interface they use is broken.

The AWS Middle East situation adds a geopolitical dimension that the AI industry hasn't had to grapple with before. If cloud infrastructure in conflict zones can take physical damage, and that damage can cascade to AI services in other regions, the concentration of AI workloads on a handful of cloud providers becomes a strategic vulnerability - not just a business risk.

Anthropic is still investigating. We'll update this article when the outage is resolved and a post-mortem is available.

Sources:

• Claude Status Page
• Claude Down and Not Working - PiunikaWeb
• Claude AI Faces Service Disruption - BusinessToday
• Anthropic's Claude Chatbot Goes Down - Bloomberg
• AWS Middle East Outage After Data Center Hit by Objects - Data Center Knowledge
• AWS Power and Connectivity Issues in Bahrain, UAE - Reuters via Investing.com
• Amazon Cloud Unit Flags Issues at UAE, Bahrain Data Centers - Deccan Herald