AI
OpenAI, Anthropic, and xAI went down Thursday—but nobody's explaining why
Three leading AI companies experienced simultaneous service disruptions on the same morning, yet both OpenAI and Anthropic have stayed silent on the root cause.
WIRED — AI
On Thursday morning, ChatGPT, Claude, and Grok all went offline around the same time—a rare coincidence that left users wondering what went wrong. While SpaceX disclosed that xAI's Grok outage stemmed from a Memphis data center failure, OpenAI and Anthropic have offered no explanation for their own service disruptions, leaving observers to speculate whether the incidents shared a common cause.
What happened#
Each platform reported problems within an hour of one another, starting before 7 am PT:
- xAI's Grok began experiencing issues at 6:30 am PT due to "an outage at our Memphis compute center," according to SpaceX. Service was restored by 10:05 am PT.
- Anthropic's Claude reported a partial outage beginning at 6:23 am PT affecting Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5. The company identified a fix and declared the issue resolved by 9:16 am PT, though Claude Sonnet 5 briefly experienced problems afterward.
- OpenAI's ChatGPT experienced a routing error at 7:43 am PT that blocked access for some users. The company deployed a solution by 8:17 am PT.
Reports also surfaced of a possible Google Gemini outage, though Google has not confirmed any incident.
The mystery of shared causes#
When multiple major AI services go down simultaneously, the usual culprit is a shared infrastructure provider—a cloud service like AWS or Azure, a content delivery network, or another backbone vendor. Notably, none of the major internet infrastructure companies reported outages on Thursday morning.
OpenAI and Anthropic have not publicly attributed their problems to any external provider or third-party failure. A spokesperson for OpenAI characterized the issue as a "routing error" but did not elaborate. Anthropic declined to comment entirely. SpaceX alone offered transparency about its internal hardware issue.
A tangled web of partnerships#
The timing raises questions about whether the companies use shared compute resources. Anthropic and xAI announced a compute partnership with SpaceX in May, suggesting potential infrastructure overlap. SpaceX added in its public statement that it wanted "to apologize to our impacted compute partners," hinting at wider consequences but providing no further details.
