Google explains cause of global internet outage

Last night around 9:00 p.m., users across the globe experienced a widespread internet outage.

The disruption affected Gmail, YouTube, Spotify, Steam, Twitch, ChatGPT, Discord, Snapchat, and dozens of online games. Reports peaked on Downdetector around 10:30 p.m.; after a brief recovery at 10:40 p.m., issues returned and weren’t fully resolved until close to midnight.

According to an investigation by Cloudflare and Google, the outage was linked to Google Cloud’s infrastructure, where an automated quota update (limiting request volumes) triggered a failure in the API management system of the global Identity & Access Management (IAM) service.

• IAM is used as a secure user access verification service. Due to the faulty update, the request quota was exceeded, leading IAM to reject external traffic en masse with 503/401 errors — causing failures across major Google products and its partners.
• Another major infrastructure provider, Cloudflare, was also affected. In a blog post, the company apologized and pointed to a “third-party vendor” as the source of the issue — which, under the circumstances and based on previous disclosures, was clearly Google Cloud.

• Most of Cloudflare Workers KV key storage is hosted on Google Cloud. KV stores configuration data for security services like Access, WARP, Gateway, Turnstile, and Stream. When IAM stopped issuing tokens to access this storage, 90% of KV requests resulted in 500/503 errors. All authorization attempts via Cloudflare Access were rejected, impacting a vast number of Cloudflare clients.

Following the incident, Cloudflare announced plans to expedite the migration of KV to its own R2 storage system to reduce dependency on third-party infrastructure — a move that had been in the works for some time.

Amid the wave of outage reports, some users also flagged issues with Amazon Web Services. However, an AWS spokesperson told TechCrunch that its infrastructure “operated normally with no reported disruptions.” Still, due to the confusion caused by Cloudflare’s outage, false complaints about AWS service failures appeared on tracking platforms, which Amazon promptly denied.