The three largest US frontier labs — OpenAI, Anthropic, and xAI — all suffered chatbot outages within roughly 90 minutes of each other on the morning of September 3, 2026, and only one of them has offered a concrete cause. ChatGPT and Codex went dark at 7:43 am PT and came back at 8:17 am PT, a 34-minute window OpenAI attributed to a routing error. Anthropic had already been logging elevated errors on Claude since 6:23 am PT, and xAI's Grok went down at 6:30 am PT.
SpaceX, xAI's parent, is the only company that pointed at physical infrastructure. In a public statement Thursday afternoon, SpaceX said the Grok downtime stemmed from "an outage at our Memphis compute center this morning," and added an apology to "our impacted compute partners." Anthropic and xAI announced a compute partnership with SpaceX in May, which makes the wording pointed: SpaceX is signaling that at least one of its AI customers was affected by the same Memphis event.
OpenAI, for its part, did not tie its downtime to any external provider. Spokesperson Kathleen Chaykowski told reporters that "a routing error starting around 7:43 am PT on Thursday, September 3, made ChatGPT and Codex unavailable for some users across platforms," and that a fix was implemented by 8:17 am PT. Anthropic declined to comment beyond its status page, which described a "partial outage" hitting Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5. The company said it "identified the cause" and marked the incident resolved at 9:16 am PT. Claude Sonnet 5 appeared to hit similar trouble briefly after 9 am PT.
Key facts
- 01OpenAI, Anthropic, and xAI all reported outages within a 90-minute window on the morning of September 3, 2026.
- 02OpenAI blamed a routing error that began at 7:43 am PT and was resolved by 8:17 am PT, affecting ChatGPT and Codex.
- 03Anthropic flagged a partial outage at 6:23 am PT affecting Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5; resolved by 9:16 am PT.
- 04xAI's Grok went down at 6:30 am PT and was restored at 10:05 am PT after SpaceX cited a Memphis compute center outage.
- 05Cloudflare, Amazon Web Services, and Microsoft Azure reported no outages on Thursday, ruling out the obvious shared causes.
The obvious explanation for three simultaneous outages at three unrelated labs is a shared upstream dependency — a hyperscaler, a CDN, or a DNS provider having a bad morning. That explanation does not hold up here. Cloudflare, Amazon Web Services, and Microsoft Azure all reported clean status boards on Thursday. No major internet infrastructure provider has claimed a Thursday morning incident that would explain the cluster.
There were scattered user reports of Google Gemini problems in the same window, but Google did not confirm anything on its status dashboard and did not respond to press inquiries. That leaves the question of whether Gemini was actually affected, or whether users simply piled onto Gemini when the other three chatbots stopped working and misread the load spike as an outage.
The SpaceX Memphis explanation partially covers xAI, and possibly Anthropic through the May compute deal. It does not obviously cover OpenAI, which runs primarily on Microsoft Azure infrastructure and has no announced SpaceX relationship. OpenAI's "routing error" framing is consistent with an internal networking configuration problem rather than a shared external failure, but the company has not elaborated on where the routing broke or which upstream systems were involved.
For enterprise customers, the practical takeaway is that the three chatbot APIs treated as substitutes at the workflow layer went down together anyway. Any redundancy strategy that assumed "if ChatGPT is out, fail over to Claude" would have failed on Thursday morning. That is not necessarily evidence of a shared cause — coincidences happen — but it is the kind of coincidence that makes procurement teams ask harder questions about shared physical dependencies among model providers.
Coverage this month has repeatedly returned to the same theme: ChatGPT, Claude, and Grok going down within 90 minutes of each other is now a documented pattern rather than a one-time event. Without a named shared cause, the market is left to guess whether Thursday reflects a genuine common-mode failure across a compute partner none of the labs will name, three unrelated incidents that happened to coincide, or something in between. The labs' willingness to leave that question open, while the outage costs them API revenue and enterprise trust, is itself the interesting data point — the specifics of what broke may be more embarrassing than the outage itself.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
