Simultaneous Outages Hit ChatGPT, Claude and Grok With No Cause Given

Three frontier chatbots went down together

On the morning of Thursday, September 3, the flagship chatbots from OpenAI, Anthropic, and xAI all suffered rare service disruptions at nearly the same time, leaving users of ChatGPT, Claude, and Grok with errors or unavailable assistants, according to WIRED.

Anthropic was the first to signal trouble. The company began reporting a partial outage at 6:23 am PT, tied to heightened error rates on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5. It later said it had found the cause and shipped a fix, and marked the incident resolved by 9:16 am PT. Claude Sonnet 5 briefly showed similar problems just after 9 am.

xAI followed at 6:30 am PT, posting on its status page that it was looking into an outage affecting Grok across all of its platforms and services. The episode was declared over at 10:05 am PT, with the company saying traffic had returned to normal.

OpenAI's disruption began later and was resolved fastest. Spokesperson Kathleen Chaykowski told WIRED that a routing error starting around 7:43 am PT left ChatGPT and Codex unavailable for some users across platforms, and that a fix was in place by roughly 8:17 am PT and still being monitored.

Only one company explained itself

In public comments on Thursday afternoon, SpaceX, which owns xAI, attributed the Grok problems to an outage at its Memphis compute center and apologized to compute partners who were caught up in it. That apology carries extra weight because, as WIRED notes, Anthropic and xAI announced a compute partnership with SpaceX in May, meaning at least two of the three affected companies have ties to the same compute operation. SpaceX did not respond to WIRED's request for further comment, and Anthropic declined to comment on the incident at all.

No shared cause has been confirmed

Simultaneous failures across an entire sector would normally point to a single cloud provider, content delivery network, or other vendor going down and dragging its customers along with it. WIRED reports that the timing initially looked like exactly that kind of pattern, but neither OpenAI nor Anthropic pointed to an outside provider, and major internet infrastructure players, including Cloudflare, Amazon Web Services, and Microsoft Azure, reported no outages on Thursday.

There were also scattered reports of a possible Google Gemini disruption on Thursday morning, but Google did not confirm them and logged no incidents on its service status dashboard. The company did not reply to WIRED ahead of publication.

Why it matters

When three of the most widely used AI assistants go dark within the same morning and almost no one involved will say why, everyone building on top of these platforms is left guessing about how fragile the underlying infrastructure really is. The SpaceX connection hints at a new kind of concentration risk: as frontier compute pools in fewer places, a problem at a single facility could plausibly ripple across multiple chatbots at once, though nothing confirmed on Thursday establishes that any of the outages shared a cause.

What is confirmed is the silence. For businesses that now route coding, research, and customer-facing work through these models, an unexplained stretch of downtime is a useful, if unwelcome, stress test of how much they depend on systems whose operators owe them very little information when things break.