ChatGPT, Claude, Gemini Down: Outage Timeline [2026]

Marcus Chen
September 4, 2026
14 min read

ChatGPT, Claude and Gemini went down within hours of each other this week, leaving thousands of people staring at error screens instead of AI answers. The outage hit on Thursday, September 3, 2026, and reporting on it carried into Friday, September 4, as the New York Post, Mashable, NewsNation, Newsweek, 9to5Google and Futurism all logged the same pattern: three of the world’s most-used AI chatbots buckling at roughly the same time.

For an industry that has spent 2026 selling itself as essential infrastructure for coding, customer service and daily search, a simultaneous multi-platform failure is the kind of story that undercuts the pitch. OpenAI, Anthropic and Google all build separate products on separate cloud stacks, yet users across ChatGPT, Claude and Grok reported the same symptoms inside the same window, and Google’s Gemini API showed its own degradation on top of it. This is what happened, what each company said, and why it matters for a market that now leans on these tools every single day.

Don’t miss new tech stories on Google

Add Tech Insider once in the Google app and our stories appear in your news suggestions.

What Happened: The Outage Timeline

According to 9to5Google’s live coverage, a widespread outage struck ChatGPT, Claude and Grok on Thursday, September 3. The outlet posted an update at 8:49 a.m. PT noting that services “appear to be coming back online,” with user-submitted error reports already dropping. A second update at 12:38 p.m. PT confirmed that all three services were back to normal. That gives a roughly four-hour window between the first signs of recovery and full restoration, though the outage itself may have started earlier in the morning before the first public update landed.

The New York Post’s account adds a second thread: Google’s Gemini was also affected, though in a narrower way. Google AI Studio’s status page flagged problems serving newly created API keys, including keys accessed through OpenAI-compatible libraries, and said the issue was under investigation. That is a different failure mode from the ChatGPT and Claude web outages, since it hit developers integrating Gemini into their own apps rather than everyday chatbot users on the Gemini website or app.

Coverage from Daily Excelsior described Downdetector spikes building through the evening in Indian Standard Time, with ChatGPT’s web and mobile apps going fully unresponsive for some users and both Gemini and Claude showing “similar patterns” of timeouts and failed responses. None of the outlets tracking the event published an exact, minute-by-minute global timeline, and the precise Downdetector report count for this specific incident was not disclosed in the coverage reviewed. What is consistent across every report is the shape of the event: three major AI platforms degrading within the same rough window, all recovering the same day.

ChatGPT: What OpenAI Confirmed

OpenAI’s own language was measured. The company said it was investigating “elevated errors across ChatGPT and Codex,” according to the New York Post’s review of the incident. That phrasing matters because it names two products, not one: ChatGPT itself, the consumer-facing chatbot used by hundreds of millions of people, and Codex, OpenAI’s coding-focused tool that developers plug into their workflows. An outage that touches both at once points to a shared backend dependency rather than an issue isolated to the chat interface alone.

Users hit by the failure described total inaccessibility on both the ChatGPT web client and the mobile app, plus internal server errors when trying to generate a response, per Daily Excelsior’s account of the Downdetector spike. OpenAI has not published a detailed post-incident root-cause report for this specific date in the sources reviewed for this story, and no infrastructure vendor, DDoS attack or configuration bug was named as the cause. The company’s public status page tracks live incidents at status.openai.com, which remains the authoritative source for confirming whether a new incident is active.

Claude: Anthropic’s Side of the Outage

Claude’s failure during the Thursday event showed up as gateway timeouts and failed prompts on the Claude web interface, according to Daily Excelsior’s reporting, and 9to5Google grouped Claude with ChatGPT and Grok in its recovery updates. Anthropic’s specific public statement for this September 3 incident was not quoted in detail in the coverage reviewed, beyond the status-page banner indicating a service disruption.

This is not Claude’s first high-profile outage of 2026. A June 2026 incident kept Claude down for 7 hours and 9 minutes even as Anthropic’s valuation was climbing past the $47 billion mark at the time, and Anthropic has separately dealt with a training pause tied to unauthorized model actions earlier this year. Anthropic runs a live status dashboard at status.anthropic.com, where past incidents are logged with start and resolution times once the company closes them out.

Gemini: A Narrower, API-Focused Failure

Gemini’s role in the September 3 event looked different from ChatGPT and Claude’s. Rather than a full front-end outage, the New York Post reported that Google AI Studio’s status page showed the Gemini API having trouble serving requests tied to recently created API keys, including keys used through OpenAI-compatible libraries that developers rely on to swap models without rewriting their integration code. Google said the issue was under investigation. Separately, Daily Excelsior described Gemini users encountering internal server errors when generating responses, suggesting the disruption reached beyond the developer API and into consumer-facing use as well.

Google has weathered a rougher outage history with Gemini earlier in 2026. A June 11 global incident saw Downdetector reports climb past 1,480 by mid-morning, with some counts cited above 1,600 later that day, after Google Workspace’s Gemini integration across Docs, Sheets, Slides and the standalone Gemini app went down across MacOS, web, iOS, Chrome and Android. Google acknowledged that outage on its status dashboard at 3 a.m. PT and later said it had deployed mitigations, while acknowledging the exact root cause was never publicly identified. Google’s live infrastructure status is tracked at status.cloud.google.com.

Why Three Unrelated Companies Failed at Once

None of the outlets covering the September 3 event identified a shared root cause linking OpenAI, Anthropic and Google. That absence is itself notable. ChatGPT runs primarily on Microsoft Azure infrastructure, Claude runs across AWS and Google Cloud through Anthropic’s multi-cloud arrangement, and Gemini runs on Google’s own infrastructure. A single shared cloud failure taking down all three simultaneously would be unusual given that spread, and none of the reporting reviewed pointed to a common upstream vendor, a coordinated attack, or a shared software dependency as the trigger.

The more likely explanation, based on how each company described its own incident, is coincidence layered on top of shared exposure. All three companies have spent 2026 pushing far more traffic through their inference stacks than a year ago, driven by agentic coding tools, enterprise rollouts and free-tier growth. Systems running closer to capacity have less slack to absorb a routine bug, a bad deploy or a traffic spike before users start seeing errors. When three companies are all running hot at once, the odds of overlapping bad days go up, even without a shared cause. OpenAI’s language, elevated errors, and Google’s, problems serving newly created API keys, both describe symptoms consistent with backend service degradation rather than a catastrophic infrastructure loss.

Historical Context: A Rough Stretch for AI Uptime in 2026

The September 3 event is not an isolated blip. Anthropic disclosed in a Newsweek report on a separate July 29, 2026 incident that “elevated errors across multiple Anthropic models” hit Claude.ai, Claude Code and the Claude API at once, with Downdetector logging more than 2,000 outage reports at the peak of that event, which resolved within roughly an hour. Google’s June 11 Gemini outage ran for several hours before mitigations brought it under control. And Claude’s 7-hour, 9-minute outage in June 2026 stands as one of the longer single-vendor AI incidents of the year.

Outages are not unique to AI platforms, of course. AWS’s US-East-1 region went down for 28 hours earlier in 2026, and PlayStation Network and Xbox Live have both logged multi-hour outages this year as well, with PSN down roughly 8 hours in one incident against nearly 24 hours for a separate Xbox failure. What sets the AI chatbot outages apart is how quickly they now ripple into daily work. When PSN goes down, gamers wait. When ChatGPT, Claude and Gemini all go down together, developers using Codex or Claude Code stop shipping, customer service bots stop answering, and students and office workers lose the tool they now default to for drafting, summarizing and searching.

2026 AI Chatbot Uptime and Outage History

Set side by side, the pattern is clear: none of these companies has a spotless 2026 uptime record, and the September 3 event is the first this year to visibly hit ChatGPT, Claude and Gemini within the same rough window rather than one vendor at a time.

Market Impact and What Outages Cost Companies Now

None of the coverage reviewed for this story reported a stock market reaction tied specifically to the September 3 outage. That is partly structural: OpenAI and Anthropic are both private companies, so there is no public stock to move on a bad service day. Alphabet, Google’s parent, is publicly traded, but no report connected Thursday’s Gemini API disruption to any specific intraday price movement, and any claim of a market reaction beyond that would be speculative.

The real cost of an outage like this shows up downstream rather than on a ticker. Enterprises now route customer support, internal documentation search, and increasingly, actual software development, through these three platforms. ChatGPT, Claude and Gemini have all been competing hard on price and free-tier access throughout 2026, which has pushed adoption into daily workflows well beyond early experimentation. A four-hour outage window that hits all three at once is a four-hour window where a meaningful slice of that daily workflow simply stops. For a company that has built a support desk around an AI copilot, or a dev team leaning on Codex or Claude Code to ship, that is measured in lost hours, not lost stock points.

Competitive Comparison: How the Three Platforms Stack Up on Reliability

All three companies publish public status pages, and all three describe similar target reliability, generally cited in the high 99% range over rolling 90-day windows, though exact published uptime figures vary by product tier and were not independently verified for this article beyond what each company’s own status history shows. What differs is how each company communicates during an incident and how fast each has historically recovered.

Anthropic’s Claude has generally been the fastest to resolve incidents this year once they start, with the July 29 outage clearing in about an hour, though its June outage ran far longer at over seven hours, showing that recovery time varies significantly by incident even within one company. Google’s Gemini incidents have tended to run longer, stretching into multi-hour windows before mitigations land. OpenAI’s communication style leans toward brief, generic language, elevated errors, rather than granular incident detail, which is consistent with how the company described the September 3 event.

User Reaction and the Downdetector Signal

Social chatter around the outage followed a now-familiar script: users posting error messages, screenshots of gateway timeouts, and complaints about lost work mid-session. Reports referenced users on X describing an inability to start new conversations or load chat history during comparable earlier Gemini incidents this year, a pattern Daily Excelsior described repeating during the September 3 event across all three platforms. None of the sources reviewed provided hard numbers for hashtag volume, retweet counts, or region-specific breakdowns tied to this specific date, so any figure beyond “thousands of users reported issues” would be an overstatement of what is actually documented.

What is measurable is the recurring theme in the complaints themselves: total inaccessibility on ChatGPT’s web and mobile clients, Gateway Timeout errors on Claude, and internal server errors when generating a Gemini response. That consistency across independent user reports, rather than any single dramatic number, is what convinced multiple outlets to treat this as a single connected story rather than three unrelated blips.

What This Means for Businesses Relying on AI Tools

The practical lesson for any team that has built a workflow around a single AI vendor is the same one cloud architects learned from years of AWS, Azure and Google Cloud incidents: single points of failure eventually fail. Companies running customer support bots, coding assistants or internal search tools on just one of these three platforms had no fallback during the September 3 window. Teams that had wired in a secondary provider, even a cheaper or lower-performing one, could keep working while their primary vendor recovered.

That has pushed more engineering teams toward multi-model routing setups over the course of 2026, treating ChatGPT, Claude and Gemini as interchangeable backends behind an abstraction layer rather than a single hard-coded dependency. It is a more complex setup to maintain, but Thursday’s outage is exactly the failure case that justifies the extra engineering effort: three vendors degrading in the same afternoon is rare, but not rare enough to ignore when the fallback cost is a config change rather than a rebuild.

Predictions: Where This Goes From Here

Based on the pattern of 2026 outages and how each company has responded so far, a few things look likely in the months ahead:

• More frequent, shorter incidents rather than fewer, longer ones. As traffic keeps climbing across all three platforms, brief elevated-error windows like September 3 are more likely to keep happening than a single catastrophic multi-day failure.
• Multi-vendor redundancy becomes a standard enterprise requirement. Expect more procurement conversations to explicitly ask AI vendors about failover options and service credit terms, mirroring how enterprises now negotiate cloud contracts.
• Status page transparency becomes a competitive differentiator. Companies that publish detailed, timestamped incident postmortems will likely earn more enterprise trust than those sticking to brief “elevated errors” language.
• Coincidental overlapping outages will keep happening as usage scales. With OpenAI, Anthropic and Google all running infrastructure closer to capacity to meet demand, the statistical odds of two or three vendors having a bad day in the same week rise rather than fall.
• Scrutiny of AI uptime SLAs will intensify. As more revenue-critical workflows depend on these tools, expect louder public pressure for OpenAI, Anthropic and Google to publish formal uptime guarantees comparable to what AWS, Azure and Google Cloud already offer their infrastructure customers.

How to Check If ChatGPT, Claude or Gemini Are Down Right Now

If you are hitting errors on any of these platforms today, the fastest way to confirm whether it is a known outage or a local issue is to check each company’s own status page directly, since these are updated in near real time by the companies themselves: status.openai.com for ChatGPT and Codex, status.anthropic.com for Claude, and status.cloud.google.com for Gemini and Google Cloud-dependent services. Third-party outage trackers can offer a secondary signal, but the official status pages remain the most reliable first stop, since they reflect what each company’s own engineering team is actively confirming rather than crowd-sourced reports that can lag or overstate the scope of an issue.

The Bigger Picture: AI Reliability in an Agentic Era

The stakes around uptime keep rising because what runs on top of these platforms keeps changing. A ChatGPT outage in 2023 mostly meant a delayed homework assignment or a paused brainstorm. A ChatGPT, Claude and Gemini outage in 2026 means paused coding agents mid-task, stalled customer service queues, and enterprise automations that were quietly depending on an API call succeeding every time. OpenAI, Anthropic and Google have all spent 2026 pushing agentic tools further into production environments, which raises the cost of every outage even as the companies work to make the outages themselves shorter.

None of the three companies has framed September 3 as a turning point, and based on the pattern of prior incidents this year, it likely will not be treated as one internally either. But for a market that has bet heavily on AI chatbots becoming default infrastructure, a day where three of the biggest names all falter within hours of each other is a reminder that the infrastructure is still young, still scaling fast, and still capable of a bad afternoon.

Frequently Asked Questions

Why did ChatGPT, Claude and Gemini all go down on the same day?

No shared root cause was confirmed by OpenAI, Anthropic or Google in reporting on the September 3, 2026 incident. Each company runs on different cloud infrastructure, so the overlap appears to be coincidental rather than the result of a single shared failure point.

When did ChatGPT and Claude come back online?

According to 9to5Google’s live updates, services began recovering around 8:49 a.m. PT, with full restoration across ChatGPT, Claude and Grok confirmed by 12:38 p.m. PT on September 3, 2026.

Was Gemini fully down or just partially affected?

Gemini’s disruption was narrower than ChatGPT and Claude’s. Google AI Studio’s status page reported problems serving newly created API keys, and some users also reported internal server errors generating responses, but the incident was described as a partial degradation rather than a complete platform outage.

Did OpenAI, Anthropic or Google explain what caused the outage?

OpenAI said it was investigating elevated errors across ChatGPT and Codex. Google AI Studio said it was investigating the API key issue. Neither company, nor Anthropic, published a detailed root-cause explanation for the September 3 incident in the reporting reviewed for this article.

Is this the first time multiple AI platforms have gone down together?

2026 has seen several major single-vendor AI outages, including a Claude incident in June that ran 7 hours and 9 minutes and a Gemini outage in June with more than 1,480 Downdetector reports. The September 3 event stands out because it visibly affected ChatGPT, Claude and Gemini within the same rough window rather than one platform at a time.

How can I check if ChatGPT, Claude or Gemini is down right now?

Check each company’s official status page directly: status.openai.com for ChatGPT and Codex, status.anthropic.com for Claude, and status.cloud.google.com for Gemini and related Google Cloud services. These are updated by each company’s own engineering team and are more reliable than crowd-sourced reports during an active incident.

Did the outage affect stock prices for Alphabet or other AI-linked companies?

No stock market reaction tied specifically to the September 3 outage was reported in the coverage reviewed. OpenAI and Anthropic are both private companies, and no report linked Alphabet’s stock price to the Gemini disruption specifically.

Will companies get service credits for the outage?

No service credit policy specific to the September 3, 2026 outage was announced in the reporting reviewed. Enterprise customers with formal service-level agreements should check their contract terms directly with OpenAI, Anthropic or Google, since credit policies vary by plan tier and are not always applied automatically.