Thursday morning delivered a rare synchronized stumble for the cloud AI sector. Anthropic kicked things off at 9:23 am Eastern with a partial outage affecting Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5. The company identified the cause within 15 minutes, deployed a fix, and marked the incident resolved by 12:16 pm. A separate blip hit Claude Sonnet 5 just after noon. OpenAI wasn’t far behind, reporting elevated errors across ChatGPT and Codex at 10:43 am that degraded performance until a mitigation brought things back online around 12:55 pm.
xAI’s Grok had it worse. DownDetector reports for the model jumped from fewer than 10 just before 9 am to 1,365 by 9:45 am — a spike that suggests something broke hard and fast rather than a slow trickle of errors. As of this writing, Grok still shows a user-facing error message saying the service “is experiencing issues” and that the company is working to restore it. Reports have since fallen to 273, which indicates recovery is underway but not complete.
The overlap is striking. Four major AI providers — OpenAI, Anthropic, xAI, and Google — all reported service disruptions within the same few-hour window. That’s not normal. These companies run on entirely separate infrastructure stacks, so a single point of failure is unlikely. What’s more plausible is a shared dependency: a common cloud provider, a CDN issue, or some kind of cascading failure in the underlying compute layer that multiple AI companies rely on. Neither OpenAI nor Anthropic has publicly commented on whether the incidents are related.
For users, the timing couldn’t be worse. Claude and ChatGPT have become default tools for writing, coding, and analysis across industries. When both go down simultaneously, there’s no obvious fallback — Grok was down too, and Google’s models were also reportedly having issues. The incident raises an uncomfortable question the industry hasn’t fully grappled with: what happens when the AI tools we’ve built our workflows around all fail at once? The answer, apparently, is that you wait. And refresh. And wait some more.
