ChatGPT, Claude, and Grok All Went Down the Same Morning. Nobody Is Saying Why.
Three major AI chatbots suffered outages within minutes of each other on Thursday. A shared cause seems likely, but the companies are not saying what it was.
Key points
- ChatGPT went offline for some users from roughly 7:43 am to 8:17 am PT on Thursday, September 3, according to OpenAI.
- Claude, Anthropic's chatbot, suffered a partial outage beginning at 6:23 am PT, with the issue marked resolved at 9:16 am PT.
- Grok, the chatbot made by xAI (Elon Musk's AI company), was down across all platforms from 6:30 am PT until 10:05 am PT.
- xAI blamed a hardware failure at its Memphis data centre; OpenAI cited an internal routing error; Anthropic gave no public explanation.
- No shared third-party service, such as a cloud network or internet infrastructure provider, has been identified as a common cause.
Three of the biggest names in AI all had their chatbots stumble on the same Thursday morning, and so far none of them has offered a full explanation for why it happened at the same time.
What actually went wrong?
Each company reported a different immediate cause, and none pointed a finger at a shared outside service.
OpenAI told Wired AI that "a routing error", meaning a technical misdirection of internet traffic inside its own systems, knocked ChatGPT and Codex (an AI coding tool) offline for some users starting at 7:43 am PT. The company said a fix was in place by 8:17 am PT, about 34 minutes later.
xAI, the AI company founded by Elon Musk and backed by his rocket firm SpaceX, said the problem was physical: "an outage at our Memphis compute centre." A compute centre is a building full of the specialised computer hardware that AI systems need to run. Grok was down from 6:30 am PT until 10:05 am PT, a window of roughly three and a half hours.
Anthropic said only that it had spotted "elevated errors" on requests to its Claude models (Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5) beginning at 6:23 am PT, identified a cause, and deployed a fix, with full resolution by 9:16 am PT. The company declined to say anything more.
Could this all be connected?
On the surface, yes. Three simultaneous outages in one sector almost always point to a shared piece of infrastructure, such as a cloud computing provider, a content delivery network (a system that moves data quickly around the internet), or some other behind-the-scenes vendor that many companies rely on.
Here is the problem: none of the big infrastructure players reported trouble. Cloudflare, Amazon Web Services, and Microsoft Azure, the kind of companies that could take down multiple AI services at once, all showed clean status boards on Thursday.
xAI's public statement did include an unusual line: "We'd also like to apologize to our impacted compute partners." That wording is notable because, in May, both Anthropic and xAI announced a compute partnership with SpaceX. Whether that shared arrangement links the two outages remains unclear, and neither company has confirmed it.
There were also scattered user reports of Google Gemini having trouble Thursday morning. Google neither confirmed nor denied an incident.
What does this mean for people who use these tools?
All three services have been restored. If you rely on any of these chatbots for work, the outages serve as a reminder that even the largest AI platforms can go dark without warning, and without much explanation.
Bookmark the status pages for the tools you depend on: they are often the fastest way to confirm whether a problem is on your end or theirs. When a chatbot stops responding, check the status page before spending time troubleshooting your own device or connection.
