SAN FRANCISCO — April 20, 2026 – The ChatGPT outage on April 20, 2026 was the most significant disruption to OpenAI’s flagship service in months, affecting thousands of users globally and temporarily taking down not just ChatGPT but also Codex and the API Platform. The outage began at approximately 10:05 AM ET and affected login, conversations, voice mode, image generation, and Codex simultaneously — with OpenAI upgrading the incident from “degraded performance” to a “partial outage” on its status page within the first hour. A fix was applied and OpenAI confirmed it was “monitoring the recovery” by 12:48 PM ET — approximately 90 to 100 minutes of meaningful disruption for many users.
The ChatGPT April 20 outage is a significant news story not just because of its scale but because of what it reveals about the infrastructure dependency that has quietly built up around a single AI platform. When ChatGPT goes down in 2026, it is not a minor inconvenience for a niche user base — it is a disruption to the daily workflows of hundreds of millions of users including professionals, students, developers, and enterprise teams whose operations depend on it.
Timeline of the April 20 ChatGPT Outage
| Time (ET) | Event |
|---|---|
| ~10:05 AM | Spike in Downdetector reports — first signs of disruption for users globally. Reports escalate rapidly from a few hundred to thousands within minutes. |
| ~10:30 AM | OpenAI’s status page acknowledges “degraded performance” — lists conversations, login, voice mode, and image generation as affected components. |
| ~11:13 AM | OpenAI upgrades classification to “partial outage” (red status). Confirms: “We are investigating the issue for the listed services. Impacted users are currently unable to access ChatGPT, Codex and API Platform.” |
| ~11:15–12:00 PM | Peak reports hit 8,700+ in the UK and 2,000–13,000 globally depending on source. Downdetector reports highest in UK, suggesting timing coincided with peak afternoon usage in Europe. |
| ~12:03 PM | OpenAI status page states “continuing to investigate” — limited technical detail shared publicly during incident. |
| ~12:30 PM | Downdetector reports begin declining significantly — service restoring for most users. |
| 12:48 PM | OpenAI confirms on status page: “We have applied the mitigation and are monitoring the recovery.” Status downgraded from red to amber. |
| ~1:15 PM | Reports return to normal baseline on Downdetector. Outage effectively over for the vast majority of users. |
ChatGPT outage timeline, April 20, 2026. Sources: TechRadar, Tom’s Guide, OpenAI status page, Downdetector, IsDown.
What Was Affected
The April 20 ChatGPT outage was notably broad in its scope. According to OpenAI’s status page, every major component of the platform was affected simultaneously: standard conversations, login, voice mode, image generation, Codex (OpenAI’s agentic coding tool), and the API Platform. This breadth is what distinguished it from the minor incidents that occur more frequently — which typically affect one component while others remain functional. Reports from users indicated different symptoms in different regions, which is consistent with a distributed infrastructure issue rather than a single point of failure.
The most commonly reported experience, according to a poll run by TechRadar during the outage, was inability to access old conversations — with 63% of affected respondents reporting this as their primary issue. 27% reported being unable to sign in entirely. A smaller subset could access new chats but received error messages on established conversations and projects. Notably, Codex appears to have continued functioning for some users even while the main ChatGPT interface was down — reflecting differences in the infrastructure serving the two products.
Key numbers from the April 20 ChatGPT outagePeak Downdetector reports: 13,000+ globally, 8,700+ in UK alone | Duration of meaningful disruption: approximately 90–100 minutes | Components affected: ChatGPT, Codex, API Platform, voice mode, image generation | OpenAI’s public communications: brief status updates only — no technical cause disclosed | Fix applied: 12:48 PM ET | IsDown tracking: this was OpenAI’s 1 major outage in the prior 90 days, out of 59 total incidents (58 minor)
Why This Outage Matters
In January 2023, ChatGPT had 100 million monthly users. In February 2026, it reached 900 million weekly active users. That growth has not been matched by a proportional reduction in downtime incidents — IsDown, which tracks official status page data rather than just user reports, shows OpenAI recorded 59 incidents in the 90 days before April 20, with a median resolution time of 1 hour 47 minutes. The April 20 outage is notable as the one “major” incident in that period, but the underlying pattern of frequent minor disruptions is itself significant for enterprise customers.
The timing is also notable. OpenAI simultaneously announced — or had recently announced — a record $122 billion funding round at an $852 billion valuation. The contrast between that headline and the service disruption experienced by thousands of users on the same day illustrates the gap between AI investment narrative and AI infrastructure reliability that enterprise procurement teams are increasingly noting in their evaluations.
⚠️ Enterprise teams: What this means for AI single-vendor dependencyThe April 20 outage is a practical demonstration of the risk of building workflows around a single AI platform. As Anthropic’s Thunderbolt launch (April 16) noted, organisations that route all AI work through one vendor’s infrastructure have no fallback when that infrastructure experiences issues. Enterprise teams using ChatGPT should ensure they have documented contingency procedures — including alternative tools and offline capability — for the minority of time when any AI platform experiences degraded service.
Click Here to Read More About AI News!
Frequently Asked Questions
What caused the ChatGPT outage on April 20 2026?
OpenAI has not publicly disclosed the technical root cause of the April 20 outage. During the incident, its status page communications were limited to “we are investigating” and “we have applied the mitigation.” The breadth of affected components — conversations, login, voice mode, image generation, Codex, and the API Platform all simultaneously — suggests a distributed infrastructure issue rather than a single component failure. OpenAI’s post-incident transparency has historically been limited; no detailed post-mortem has been published at the time of writing.
How long was ChatGPT down on April 20 2026?
Approximately 90 to 100 minutes of meaningful disruption for most affected users. The outage began around 10:05 AM ET, OpenAI confirmed a fix was applied at 12:48 PM ET, and Downdetector reports returned to baseline levels by approximately 1:15 PM ET. The experience varied significantly by region and user — some users in the same country could access new chats while others could not log in at all, reflecting the uneven nature of the outage across distributed infrastructure.
How many people were affected by the ChatGPT outage?
Downdetector recorded over 13,000 simultaneous outage reports globally at peak, with over 8,700 reports in the UK alone. Given that Downdetector captures a fraction of actual affected users — most people experiencing issues do not report to the platform — the actual number of affected users was substantially larger. With ChatGPT at 900 million weekly active users globally, even a partial outage affecting a small percentage represents millions of disrupted sessions.
