Nobody Is Saying Why OpenAI and Anthropic Had Outages Today
Major AI companies OpenAI and Anthropic experienced rare, simultaneous outages on Thursday morning, affecting their popular chatbots, while xAI's Grok also went down. Despite the coincidence, neither OpenAI nor Anthropic attributed their issues to a shared third-party pro…
Intelligence analysis by Gemini 2.5 Flash
On Thursday, several leading AI models, including ChatGPT, Claude, and Grok, suffered concurrent outages, sparking speculation about a common underlying cause. While xAI's parent company, SpaceX, cited a compute center issue and apologized to partners, OpenAI reported a routing error, and Anthropic declined to comment, leaving the broader reason for the widespread disruption unclear.
Imagine if all the big toy factories that make your favorite robots suddenly stopped working at the same time, but nobody would say exactly why. One factory, xAI, said its power went out in one of its buildings. But the other two big factories, OpenAI and Anthropic, just said they fixed their own problems without explaining if they were connected or if something big happened to everyone at once. It makes you wonder if they all use the same special machine, or if something else big happened that they're not telling us.
Analysis
The synchronized downtime experienced by frontier AI models from OpenAI, Anthropic, and xAI on Thursday morning presented a puzzling scenario for the technology sector. While such concurrent issues often point to a shared dependency, like a major cloud provider or content delivery network, the responses from the affected companies offered disparate explanations, deepening the mystery surrounding the incident.
SpaceX
SpaceX, the parent company of xAI, was the only entity to offer a specific, albeit brief, explanation for its Grok chatbot's outage. The company publicly stated that the issues stemmed from "an outage at our Memphis compute center this morning." This direct acknowledgment provided some clarity regarding Grok's specific disruption, contrasting sharply with the more guarded statements from its peers.
Intriguingly, SpaceX also included an apology to "impacted compute partners," a detail that could be significant given Anthropic and xAI's announced "compute partnership" with SpaceX in May. This suggests a potential link, at least for Anthropic, though Anthropic itself did not confirm this connection or elaborate on its own outage's cause. The incident highlights the complex web of infrastructure and partnerships that underpin modern AI services.
OpenAI
OpenAI, the developer of ChatGPT and Codex, provided a more technical, yet still isolated, explanation for its service interruption. A spokesperson, Kathleen Chaykowski, informed WIRED that the outage was caused by "a routing error starting around 7:43 am PT on Thursday, September 3." The company quickly implemented a solution, with services reportedly restored by 8:17 am PT, and continued monitoring the fix.
OpenAI's explanation of a routing error suggests an internal network issue rather than an external dependency. This contrasts with the broader speculation of a shared third-party problem. The swift resolution indicates a degree of internal control over their infrastructure, but the initial failure still points to the inherent complexities and potential points of failure within large-scale AI systems. The lack of any mention of external partners or shared infrastructure in their statement further isolates their incident from the others, at least from their perspective.
Anthropic
Anthropic, known for its Claude models, maintained a notably tight-lipped stance regarding its outage. The company initially alerted users to a "partial outage" involving "elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5" at 6:23 am PT. Shortly thereafter, Anthropic claimed to have "identified the cause" and deployed a fix, marking the issue as resolved by 9:16 am PT.
Despite the resolution, Anthropic "declined to comment on the episode" when approached by WIRED. This silence, especially in the context of simultaneous outages across the sector and its known partnership with SpaceX, fuels speculation. The company's decision not to elaborate on the cause, or to confirm or deny any shared infrastructure issues, leaves a significant gap in understanding the broader implications of Thursday's events for the AI industry's stability and interdependencies.
Key points
- OpenAI, Anthropic, and xAI experienced rare, simultaneous outages of their frontier AI models on Thursday morning.
- SpaceX, xAI's parent company, attributed Grok's outage to an issue at its Memphis compute center and apologized to 'impacted compute partners'.
- OpenAI stated its outage was due to a 'routing error' and was resolved within an hour.
- Anthropic reported a 'partial outage' with 'elevated errors' on its Claude models, identified a cause, and deployed a fix, but declined further comment.
- Despite the concurrent nature, neither OpenAI nor Anthropic cited a shared third-party service provider as the cause, leaving the broader reason for the widespread disruption unclear.
The simultaneous outages could prompt major AI developers to invest more heavily in redundant and diversified infrastructure, leading to more robust and reliable AI services in the long run. This increased focus on resilience might also foster greater transparency about system dependencies and potential vulnerabilities, ultimately strengthening the entire AI ecosystem.
The lack of transparency surrounding these concurrent outages raises concerns about the stability and interconnectedness of critical AI infrastructure. If a shared, undisclosed vulnerability or single point of failure exists, it could lead to more widespread and impactful disruptions in the future, potentially undermining trust in frontier AI models and their reliability.
