OpenAI Suffered a Near-Global Outage Last Night — and Gave Every Codex and ChatGPT Work User a Free Reset
OpenAI’s infrastructure went down hard in the early hours of July 25, 2026. Between roughly 2am and 4am, the company suffered what its own team described as an “almost global outage” — hitting APIs, ChatGPT, and Codex simultaneously. By morning, everything was back up. And then came the goodwill gesture.
Tibo (@thsottiaux), who leads engineering at OpenAI, confirmed the incident publicly and announced that the company had reset usage limits across all Codex and ChatGPT Work accounts. “We learn. We reset. Enjoy,” he wrote — a short post that racked up 849K views within hours.
We have reset usage limits for all Codex and ChatGPT Work users.
Last night around 2am to 4am we suffered an almost global outage. All well and recovered, but you know what comes next.
We learn. We reset. Enjoy.
— Tibo (@thsottiaux) July 26, 2026
What the OpenAI Outage Actually Affected
The outage hit three of OpenAI’s most-used surfaces at once: its core APIs, ChatGPT, and Codex. The company’s status page showed elevated error rates across all three, with the incident listed as ongoing for at least 17 minutes before recovery efforts took hold. A two-hour window of degraded or unavailable service is significant at OpenAI’s current scale.
For context, OpenAI crossed 8 million active users on Codex and ChatGPT Work in mid-July. A two-hour outage across that user base during working hours in some time zones isn’t a minor hiccup — it’s a meaningful service disruption, and the company treated it accordingly.
Why a Usage Reset Is the Right Response
The decision to reset limits, rather than simply apologise, reflects a maturing approach to incident recovery. Usage-limited products like Codex operate on weekly or rolling quotas. When the service goes down mid-cycle, users lose time they paid for — and a written apology doesn’t restore that.
A full reset gives every affected user their quota back in full, regardless of how much they had already consumed. It’s a clean, operationally simple move that converts a bad experience into an unexpected benefit. The response in comments was immediate: one user replied that they had just burned through their limits and were left with 8% remaining — they got it all back.
Furthermore, this isn’t the first time OpenAI has used limit adjustments as a user relations tool. The company lifted the 5-hour usage cap entirely earlier this month — a separate move, but part of the same pattern: OpenAI is increasingly treating usage flexibility as something it can deploy strategically, both to reward users and to smooth over rough patches.
The Infrastructure Question Underneath This
An almost-global outage affecting APIs, ChatGPT, and Codex simultaneously points to a shared dependency failure somewhere in the stack — not an isolated service bug. That’s worth noting, because OpenAI has been operating under surging demand since GPT-5.6 Sol launched and Build Week drove a sharp spike in Codex usage.
Inference infrastructure at this scale is genuinely hard. The company has been candid about the fact that demand has consistently outpaced its ability to provision capacity — and that pressure creates fragility. A two-hour global window, however, suggests something more acute than capacity saturation. Recovery within that window is competent execution; preventing the failure in the first place is the harder, longer-term problem OpenAI still needs to solve.
For now, users have their limits back. The outage is resolved. And OpenAI has turned a service failure into a small moment of goodwill — which, at 849K views on a single post, landed exactly as intended.
Frequently Asked Questions
When did the OpenAI outage happen?
The outage occurred between approximately 2am and 4am on July 25, 2026, affecting OpenAI’s APIs, ChatGPT, and Codex simultaneously.
Who is eligible for the Codex usage reset?
All Codex and ChatGPT Work users received a full usage limit reset, according to the announcement by OpenAI’s Tibo (@thsottiaux) on July 26, 2026.
Did the OpenAI outage affect the API?
Yes. The outage hit OpenAI’s core APIs, ChatGPT, and Codex at the same time, with elevated error rates confirmed on the company’s status page.
How long did the OpenAI outage last?
The incident lasted approximately two hours, from around 2am to 4am on July 25, 2026.
