ChatGPT outage reports jumped from 5,000 to more than 22,000 in 10 minutes on Thursday morning. Claude, Grok and coding tool Cursor were already broken.

Three rival labs failed inside roughly 90 minutes, and no cloud provider declared a fault. Cardano founder Charles Hoskinson says it looked like a nation state, naming Gemini as the only exception as it uses its own TPUs.

How the AI Outage Happened

Grok and Claude users began reporting errors near 9 a.m. Eastern Time (ET). ChatGPT complaints spiked 90 minutes later, Downdetector data showed.

OpenAI was the only lab to confirm anything, logging elevated errors across ChatGPT and Codex, its coding agent.

Anthropic users hit a capacity wall instead.

“Due to unexpected capacity constraints, Claude is unable to respond to your message. Try again soon,” read Claude in-app error message.

xAI declared no incident, yet Grok told users its model was unavailable.

Follow us on X to get the latest news as it happens

Charles Hoskinson Floats a Nation-State Theory

The Cardano executive, who also co-founded Ethereum, says it looks like a nation state hit the three AI models at once, citing their reliance on Nvidia.

“It looks like a national state brought down Claude, ChatGPT, and Grok” Hoskinson wrote.

This means he thinks a government, not a hacker or a technical glitch, deliberately knocked the three chatbots offline.

Google logged no Gemini incident, yet users filed hundreds of reports and it flagged a fault serving new API keys. A shared chip supplier is not a shared failure path. Nvidia hardware in different buildings does not fail together.

The damage was uneven, with Claude and Grok peaking near 1,500 reports each while ChatGPT cleared 35,000.

No lab has reported an attack, but OpenAI listed 19 affected components and Anthropic blamed its Opus models.

The incident has turned it into a case for decentralized AI, one plugging Midnight, the privacy sidechain from Hoskinson’s Cardano ecosystem.

Shared Compute Draws Scrutiny

Attention turned to one building in Memphis. SpaceX merged with xAI in February, then rented Anthropic the full Colossus 1 compute capacity, built for Grok.

Anthropic draws over 300 megawatts across 220,000 Nvidia chips there, just under half of xAI’s roughly 500,000-GPU fleet.

Two of Thursday’s casualties sit on the same racks. However, neither firm has linked that to the failures.

“Grok, Claude, and ChatGPT all utilize significant portions of Colossus 1 in Memphis. That is the likely source of the outage… The question is whether it was accidental or sabotage,” one user posed.

A Cloudflare outage hit crypto platforms last November, with two AWS failures knocking services offline weeks earlier.

Anthropic went dark alone in a similar outage in March.

Concentration stays cheap until everything stops at once.

The outage comes as multiple AI models debut within days of each other.

Source: BeInCrypto