Several of the world's largest artificial intelligence (AI) services registered interruptions on the 3rd, leading to speculation about a common cause, even though the responsible companies are distinct.
ChatGPT, Claude, and Grok experienced problems at the same time, despite being operated by OpenAI, Anthropic, and SpaceXAI, respectively. Among the notable chatbots, only Gemini appears to have remained stable, although it also registered brief spikes in failure reports on Downdetector during the same period as its competitors.
The coinciding occurrence raised questions about a potential shared reason, but so far, no official explanation links the three incidents. Hypotheses raised include failures in infrastructure services or issues with computational capacity supply.
OpenAI was the only one of the three companies to respond to journalists' inquiries. In a statement sent to various media outlets, a spokesperson informed that a routing error began around 7:43 AM Pacific Time (equivalent to 11:43 AM Brasília time) on Thursday (3rd), resulting in the unavailability of ChatGPT and Codex for certain users across several platforms.
According to the company, a fix was applied around 8:17 AM (12:17 PM Brasília time), and service monitoring continued. OpenAI did not establish any link between this issue and the outages recorded in the other two companies' services.
Anthropic, for its part, communicated that only the Opus 4.8 and Opus 5 models were affected. SpaceXAI reported a more comprehensive interruption in its services.
One of the first theories to justify the synchronized drop pointed to Microsoft Azure. As reported by 9to5Google, the platform registered an increase in complaints during the same interval when ChatGPT, Claude, and Grok experienced failures. This fueled suspicion that a cloud infrastructure failure might be linked to the events, but this theory was not confirmed, as the platform did not indicate errors in its servers.
Cloudflare was also considered a possible source of the problems. Its CTO, Dane Knecht, posted on X that the service was not experiencing any disruption. The company, which provides infrastructure services to various technology corporations, was questioned by The Register about a possible connection to the outages. The response was negative, with a spokesperson stating that Cloudflare was not facing significant interruptions and that its services were operating normally. The same publication noted that Azure did not appear to have technical difficulties at that time.
Another clue came from SpaceXAI itself. The company, previously known as xAI before being integrated into Elon Musk's rocket company, operates large data centers located near Memphis. In addition to using this capacity for its own services, the company also provides computing for other AIs. Anthropic is one of these partners, having both announced a computing collaboration in May.
On Thursday, SpaceXAI admitted that a failure in its Memphis data center compromised the operation of Grok. In a social media post, the company apologized for the inconvenience caused to both users and affected computing partners. This incident offers a plausible explanation for the drop in Grok and for Anthropic's services, given that SpaceXAI itself recognized the impact on partners and Anthropic has a computing agreement with it.
However, the case of ChatGPT remains isolated. While OpenAI attributed its failure to a routing error, the problem recognized by SpaceXAI occurred specifically in its Memphis data center. For now, there is insufficient evidence to conclude that the three shutdowns had a single origin, and the coincidence of the interruptions lacks a definitive explanation.