The outages hit fast and wide on Thursday. Anthropic, OpenAI, and xAI all reported problems with their chatbots in the same morning window, and nobody is saying why OpenAI and Anthropic had outages today.
The three companies run some of the most-used AI models in the world. When they go down at once, the instinct is to look for a shared cause. So far, the companies are not pointing at one.
The Grok Outage and the Memphis Compute Center
xAI was the first to name a culprit. SpaceX, xAI’s parent company, said on Thursday afternoon that the issues with Grok resulted from “an outage at our Memphis compute center this morning.”
That is a specific, physical explanation. The company also apologized to partners. “We’d also like to apologize to our impacted compute partners,” SpaceX said as part of its public comments on Thursday.
The apology matters because of a business tie. Anthropic and xAI announced a “compute partnership” with SpaceX in May. That link raised the question of whether one failure could ripple across both companies.
SpaceX did not respond to WIRED’s request for comment.
OpenAI Blames a Routing Error
OpenAI gave a different explanation. Spokesperson Kathleen Chaykowski told WIRED that the problem was not hardware but traffic direction.
“A routing error starting around 7:43 am PT on Thursday, September 3, made ChatGPT and Codex unavailable for some users across platforms,” Chaykowski said. “As of about 8:17 am PT on Thursday, a solution was successfully implemented and is continuing to be monitored.”
That timeline puts OpenAI’s outage entirely inside the morning window. It started at 7:43 am PT and was fixed by 8:17 am PT.
OpenAI did not mention any external vendor or shared infrastructure. The company attributed the event to its own routing.
Anthropic’s Partial Outage and the Claude Models
Anthropic declined to comment on the episode. The company did, however, post a public trail of status updates.
The company began alerting about a “partial outage” at 6:23 am PT on Thursday. The issue involved “elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5.”
Three Claude models were affected. Shortly after the alert, the company said it had “identified the cause” and that “a fix has been deployed.”
The company marked the issue as resolved by 9:16 am PT.
There was a brief follow-up. Claude Sonnet 5 seemed to have similar issues shortly after 9 am PT. That second blip appears to have been short-lived.
Anthropic never named an external cause. The company did not say whether the problem was its own or shared.
The Timeline of a Chaotic Morning
The outages overlapped but did not match exactly. Here is how the morning broke down by company:
| Company | Service Affected | Outage Start | Resolution | Stated Cause |
|---|---|---|---|---|
| Anthropic | Claude Mythos 5.1, Claude Fable 5.1, Claude Opus 5 | 6:23 am PT | 9:16 am PT | Identified cause, no external source named |
| xAI | Grok, all platforms | 6:30 am PT | 10:05 am PT | Memphis compute center outage |
| OpenAI | ChatGPT, Codex | 7:43 am PT | 8:17 am PT | Routing error |
| Google (unconfirmed) | Gemini | Reports only | No confirmation | No incident recorded |
Anthropic started first. xAI followed within minutes. OpenAI came more than an hour later.
The issues initially appeared to be linked because they coincided. A shared third-party service provider seemed like a possible explanation, but neither OpenAI nor Anthropic cited an external source in comments to WIRED on Thursday.
xAI’s Public Status Page
xAI reported Grok outages across all of its platforms and services beginning at 6:30 am PT. That is when the company posted “investigating outage” on its service status page.
The message was direct. “Grok is experiencing issues. We are working on restoring service as quickly as possible,” the page said.
The episode ran longer than the others. At 10:05 am PT the episode was marked complete.
“We have resolved the situation, and traffic is healthy again,” the company wrote.
That is about three and a half hours from first alert to all-clear. It is the longest outage of the morning.
The Unconfirmed Google Gemini Reports
There were scattered reports of a possible Google Gemini outage on Thursday morning as well.
Google did not confirm this. The company did not record any incidents on its service status dashboard.
Google did not respond to WIRED’s request for comment ahead of publication.
That leaves Gemini as a rumor, not a fact. The reports exist, but the company has not acknowledged them.
Looking for a Shared Cause
When multiple companies in the same sector fail at once, the standard explanation is a third-party vendor. Cloud providers, content delivery networks, and other infrastructure services sit beneath many products.
A single failure there can take down multiple customers at once.
But that explanation does not fit here. OpenAI and Anthropic did not point to a potential shared cause.
The major players in the internet infrastructure space did not report outages on Thursday. Cloudflare, Amazon Web Services, and Microsoft Azure all stayed quiet.
None of them logged incidents. That leaves the shared-cause theory without a confirmed trigger.
What the Companies Are Not Saying
The most notable fact is the silence.
Anthropic declined to comment. OpenAI gave a routing error explanation. xAI blamed its own compute center.
None of them blamed a common vendor. None of them pointed at each other.
The timing remains suspicious. Three major AI providers, three outages, one morning.
The companies may simply have had independent failures. That happens. Infrastructure is complex, and systems break on their own.
But the coincidence is striking enough that WIRED asked directly. The answers did not resolve the question.
Nobody is saying why OpenAI and Anthropic had outages today.
The Compute Partnership Angle
The May partnership between Anthropic and xAI adds another layer. Both companies work with SpaceX on compute.
If SpaceX’s Memphis center failed, it could plausibly affect both. xAI said the Memphis outage caused the Grok problems.
Anthropic did not say whether its own outage came from the same place. The company has a compute partnership with SpaceX, but it did not cite that as a cause.
The shared-infrastructure theory is possible but unproven. No company has confirmed it.
What Happens Next
The outages are resolved. Traffic is healthy again, at least according to the companies.
The longer question is whether these events were connected. That answer has not come.
Anthropic and OpenAI may release more details. They may not. The companies are under no obligation to explain beyond their status pages.
For users, the practical takeaway is simple. AI services can fail, and they can fail at the same time.
The systems came back within hours. The mystery of the shared morning remains open.
Source: wired.com
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.