Analysis
OpenAI, Anthropic and SpaceXAI were all hit by service outages within hours of each other Thursday, disrupting three of the largest AI model providers in the US at nearly the same time, Bloomberg reported. OpenAI said a routing error beginning around 7:43am PT made ChatGPT and Codex unavailable for some users across platforms; The Register put the plainest framing on it: ChatGPT, Claude and Grok all went down at once.
OpenAI's status page showed the outage running from roughly 10:58am to 4:55pm ET -- about six hours -- affecting ChatGPT conversations, login, search, file uploads, Voice mode, GPTs, image generation, Deep Research, Agent and ChatGPT Work. Downdetector reports peaked above 37,000 around 11am ET before falling below 1,500 by noon. Codex remote-control users needed to re-pair their mobile devices after the incident cleared.
Three Separate Stacks, One Bad Day
- OpenAI -- ChatGPT and Codex both affected by a routing error; roughly six-hour incident, resolved by mid-afternoon ET. Competitors: Anthropic, Google Gemini.
- Anthropic -- Claude's Opus 4.8 and Opus 5 models continued showing errors after other models recovered, per company statements. Pulse has tracked Anthropic's infrastructure buildout all year, including tens of billions of dollars in new compute leases.
- SpaceXAI -- the rebranded compute arm running Grok, traced its outage to a Memphis data center and apologized to Grok users and "compute partners" for a disruption lasting more than three hours.
Downdetector spikes of this size are not new for any single provider -- OpenAI alone has weathered multiple five-figure-report outages in 2026 -- but three separately operated infrastructure stacks degrading inside the same few-hour window, without one company's postmortem naming the other two, is a different pattern than the isolated incidents each has weathered individually before.
For any startup with a single-provider dependency baked into its product, Thursday is the concrete argument for a fallback API path -- not as an abstract resilience exercise, but because three of the four largest US frontier labs went down inside the same trading day. Enterprise customers evaluating AI vendor contracts now have a fresh, dated incident to cite in negotiating SLA terms and multi-provider redundancy clauses.
There's a simpler read that undercuts any shared-cause theory: no evidence has actually established one. OpenAI's own explanation (a routing error), SpaceXAI's (a Memphis data-center issue) and Anthropic's (unspecified, tied to specific newer models) are three different technical stories, not one. Reading a single infrastructure failure into three unrelated incidents overstates what's actually been confirmed, and the more mundane explanation -- coincidence across an industry running enormous, still-immature compute stacks -- fits the evidence just as well.
OpenAI, Anthropic and SpaceXAI have each published only partial explanations so far, and none has said whether the timing was coincidental -- the next credible signal is whether any of the three publishes a full postmortem naming a shared vendor, cloud region or upstream dependency.
