ChatGPT, Claude and Grok all broke the same morning. Only one company said why
On Thursday, September 3, three of the biggest AI chatbots went down within hours of each other. One company named a cause, one gave a partial answer, and one said nothing at all.
If your chatbot froze on Thursday morning, it wasn't your Wi-Fi. Anthropic started flagging a "partial outage" at 6:23 am PT, xAI posted "investigating outage" for Grok at 6:30 am PT, and OpenAI says a routing error made ChatGPT and Codex unavailable for some users starting around 7:43 am PT. Wired called these rare outages, and they were: three separate companies, one morning.
Then came the explanations, or the lack of them. SpaceX, which owns xAI, said Grok's problems came 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. OpenAI spokesperson Kathleen Chaykowski said a solution was implemented by about 8:17 am PT and is still being monitored. Anthropic said it had "identified the cause" and deployed a fix, marked the incident resolved at 9:16 am PT, and declined to comment further.
When several companies fall over at the same time, the usual explanation is a shared supplier — the electrical substation behind the whole street, not three separate blown fuses. But Cloudflare, Amazon Web Services and Microsoft Azure all reported no outages that day, and neither OpenAI nor Anthropic pointed to any outside cause. So either it was a coincidence, or the shared piece is something none of them has named.
The tools are now everywhere. The explanations are still optional.
That gap matters more than it sounds. You can read a public status page for your bank, your electricity, your phone network. Here you get a timestamp, the phrase "a fix has been deployed," and nothing about what broke. There were also scattered reports of a Google Gemini outage the same morning — Google logged no incident on its status dashboard and did not respond to Wired. If a fourth one wobbled and no record exists, the honest summary is that we don't know how connected any of this is.
Anyone who now does part of their job inside a chat window: the 24-year-old writing code with Codex, the freelancer whose entire client work sits in Claude, the student two hours from a deadline. When the box goes grey, there's no help desk and no explanation — just refresh. And anyone building a small business on top of these tools, paying monthly like a phone plan, without a contract that entitles them to know why the service stopped.
Nobody has promised an investigation, a report or a deadline. OpenAI said its fix "is continuing to be monitored," xAI wrote at 10:05 am PT that "traffic is healthy again," and Anthropic declined to comment. That's the whole public record. What actually caused three simultaneous failures, and whether it was one cause or three, nobody has said — and no company has said it will.
The most revealing line of the day wasn't about downtime. It was SpaceX apologising "to our impacted compute partners" — a hint that rival labs may be renting from the same walls. If that's true, the AI industry looks less like several competitors and more like several storefronts on one power line.
Sources: Wired, "Nobody Is Saying Why OpenAI and Anthropic Had Outages Today," by Lily Hay Newman, September 3, 2026.
Три главных чат-бота мира легли почти одновременно, и ни одна компания не объяснила почему — это показывает, насколько миллионы людей уже зависят от инфраструктуры, устройство которой им никто не раскрывает.
Written by THE TELL’s AI newsroom. how we work · corrections