Skip to content
IFAI

Industry

ChatGPT recovers from a major 90-minute outage

Published Reach 40

OpenAI just restored ChatGPT and its developer APIs after a major outage knocked services offline for over 90 minutes. The crash hit its 100 million weekly users with capacity errors following DevDay launch hype. Sam Altman warned that surging demand means short-term instability will likely stick around.

ChatGPT recovers from a major 90-minute outage
Image: theverge.com

OpenAI just had its biggest week ever, followed immediately by its biggest headache.

On Wednesday morning, a major outage knocked both ChatGPT and OpenAI's developer APIs completely offline for more than 90 minutes. Users attempting to access the chatbot were greeted with a blunt message: ChatGPT is at capacity right now.

Hugged to death after DevDay

The outage struck just before 9 AM ET and affected ChatGPT's 100 million weekly active users, alongside over 2 million developers building on OpenAI's platform. It came less than 24 hours after a separate partial outage hit the services on Tuesday night.

OpenAI CEO Sam Altman took to X to apologize, admitting that demand for the features unveiled at Monday's developer conference was far outpacing internal expectations.

Those announcements included GPT-4 Turbo and custom GPTs, which allow anyone to build bespoke AI assistants. Altman warned users that due to extreme load, service instability is likely to persist in the short term, signing off with a brief apology.

Before this week, ChatGPT had enjoyed months of smooth, uninterrupted uptime. But the flood of new traffic generated by DevDay clearly overwhelmed the company's infrastructure.

Why it matters

We think this 90-minute blackout is a reality check for the entire AI ecosystem.

For developers and enterprises, ChatGPT is no longer a fun novelty. It is critical infrastructure. When the API goes dark for an hour and a half, entire products, automated workflows, and customer-facing agents grind to a dead stop.

If you run a startup or business powered solely by OpenAI, this is your cue to build redundancy. Single-vendor dependency in AI is officially dangerous. Having an alternative fallback model is now table stakes.

It also highlights the brutal physical limits behind AI hype. OpenAI wants to eat the entire software stack by hosting millions of custom agents. But running frontier models at global scale requires massive computing power, and right now, the infrastructure is clearly struggling to keep up with the ambition.

Sources (1)

  1. [1]
    theverge.com(opens in a new tab)Originaltheverge.com

Topic trackerOpenAI service outages and DDoS attacks

Every story on this topic, newest first.

  1. ChatGPT Plus subscriptions are back on
  2. OpenAI pauses new ChatGPT Plus signups
  3. OpenAI fends off massive DDoS attack to finally launch GPTs
  4. ChatGPT recovers from a major 90-minute outageThis story

Best of AI, daily