Cloud Outage Analysis

OpenAI Outage Cost - ChatGPT & API History, Timeline & Your Exposure

Updated July 2026 · 4 major incidents documented

Direct Answer

OpenAI's largest documented outage was 11 December 2024, a roughly four-and-a-half-hour global failure (15:16 to 19:38 PST) that took down ChatGPT, the API, and Sora at once. Per OpenAI's own post-incident report, a new telemetry service made every node in each large Kubernetes cluster run heavy control-plane operations, which overwhelmed the Kubernetes control plane and broke DNS-based service discovery. On 10 June 2025, elevated error rates degraded ChatGPT, the API, and Sora for most of a working day. Unlike AWS, Azure, and GCP, OpenAI publishes no standard uptime SLA: a 99.9% commitment exists only on enterprise Scale Tier and Priority Processing agreements. If your product depends on the API, an OpenAI outage is a full feature outage, so a fallback model or provider is the main mitigation. Full chronology and SLA detail below.

OpenAI Outage History

DateServicesDurationImpactEst. Cost
10 Jun 2025ChatGPT, API, Sora (many components degraded at once)~10 hoursGlobal; extended elevated error rates and instability across ChatGPT, the API and Sora; systemic rather than a single component. OpenAI published a service-disruption FAQNot published
26 Dec 2024ChatGPT, API, Sora~5 hoursWidely reported as an Azure data-center power failure that cascaded to OpenAI, which runs on Azure; began around 13:30 ETNot published
11 Dec 2024ChatGPT, API, Sora (all services)~4h 22m (15:16-19:38 PST)Global; a new telemetry service overwhelmed the Kubernetes control plane and broke DNS-based service discovery. API recovery began ~17:40 PST, ChatGPT/Sora ~18:50 PSTNot published
8-9 Nov 2023ChatGPT and API (periodic)~2 days (intermittent)OpenAI attributed periodic outages to an abnormal traffic pattern consistent with a DDoS attack, in the week after DevDay; Anonymous Sudan claimed responsibilityNot published

OpenAI does not disclose customer impact figures and no analyst firm has published a firm loss estimate for these incidents, so cost is marked "Not published" rather than guessed. Dates, durations, timestamps, and root causes are taken from OpenAI's own status-page post-incident reports (the 11 December 2024 Kubernetes control-plane incident and the 10 June 2025 service-disruption FAQ) and contemporaneous reporting for the 26 December 2024 Azure power cascade and the November 2023 DDoS period. Use the calculator to translate a duration like these into your own exposure. Updated July 2026.

OpenAI SLA & Service Credits

Plan / tierUptime SLA
ChatGPT Free / Plus / ProNo published uptime SLA
API - standard (pay-as-you-go)No published uptime SLA
API - Scale Tier / Priority Processing99.9% uptime SLA with service credits, on enterprise agreements only; terms negotiated privately, not published
Compensation basisService credits per the enterprise agreement, not compensation for your business loss

Per OpenAI's published Scale Tier and Priority Processing pages. Standard consumer and API usage has no contractual uptime guarantee, so treat OpenAI as a dependency that can fail without recourse. How SLA credits compare to real loss - credits cover only a fraction of business loss.

Frequently Asked

What caused the December 11 2024 OpenAI outage?
OpenAI's largest documented outage ran 15:16 to 19:38 PST on 11 December 2024 (about 4 hours 22 minutes) and took down ChatGPT, the API, and Sora together. Per OpenAI's post-incident report, a new telemetry service made every node in each large Kubernetes cluster run heavy Kubernetes API operations, which overwhelmed the control plane and broke DNS-based service discovery. The change had passed staging tests, where the load did not appear at production scale. API traffic began recovering around 17:40 PST, ChatGPT and Sora around 18:50 PST.
What happened in the June 10 2025 OpenAI outage?
On 10 June 2025 OpenAI reported elevated error rates across ChatGPT, the API, and Sora that lasted most of a working day (roughly ten hours of degradation before full recovery). The status page showed many components degraded at once rather than one isolated failure, and OpenAI published a June 10th Service Disruption FAQ afterwards. OpenAI does not publish customer loss figures.
Does OpenAI have an uptime SLA?
Not for standard usage. Consumer ChatGPT (Free, Plus, Pro) and standard pay-as-you-go API usage carry no published uptime SLA. OpenAI offers a 99.9% uptime SLA only through Scale Tier and Priority Processing for customers on enterprise agreements, with service credits if it is missed; those terms are negotiated privately. Unlike AWS, Azure, and GCP, there is no standard public SLA, so default API usage has no contractual uptime guarantee.
How much does an OpenAI outage cost a business?
It depends on how much of your product depends on the API. For a company whose core feature is an AI assistant, an OpenAI outage is a full feature outage: the revenue and productivity lost while it is down, calculated like any IT downtime. OpenAI discloses no customer impact and no analyst firm has published a loss estimate, so event cost is Not published. Use the calculator to model your own exposure from the share of revenue and workflow that runs through the API.

Updated 2026-04-27