Will OpenAI die soon?

Why do people think OpenAI is dying in late 2026?

Because OpenAI is no longer the obvious winner. For three years, it was the biggest and the most talked-about AI company. In 2026, that changed.

That is the tension. OpenAI’s own numbers are still some of the strongest in tech: more than 1.2 billion weekly ChatGPT users in October 2026, annualized revenue near $50 billion in September according to reports, a completed $122 billion funding round at an $852 billion valuation, and ChatGPT ads at a $1 billion run-rate in August. On those numbers alone, “OpenAI is dying” sounds absurd.

But the signals around the company have darkened in 2026. Anthropic now books more reported revenue per quarter than OpenAI.

Anthropic, not OpenAI, is set to list first, with an IPO targeted for November 2026.

OpenAI no longer tops the main model leaderboards. Google and Anthropic do.

In 2026, OpenAI lost its applications chief, its former COO, its head of safety, and in October it said its compute chief would leave in November. At least eight US states sued it on October 15. And its IPO slipped to 2027.

So the real question is this: over the last 6 to 9 months, do the facts show a company still compounding, or a company losing control of the market it created?

Are people abandoning ChatGPT?

No. There is no sign that people are leaving ChatGPT. The sign is that ChatGPT now shares their attention.

The headline number is still huge. On October 7, 2026, alongside the wider GPT-6 rollout, OpenAI said more than 1.2 billion people use ChatGPT every week. That is up from more than 900 million in February and more than 1 billion at the end of August. A product that adds 300 million weekly users in eight months is not being abandoned.

ChatGPT also still has more than 50 million paying consumers. That is the last official count, from March 2026. OpenAI has not published a newer one.

The second layer is less comfortable. Similarweb estimates that ChatGPT’s share of visits to the major AI chatbot websites fell from 78% in August 2025 to 57% in August 2026. Claude’s share rose from 2% to 9.6% over the same 12 months. Gemini took much of the rest.

Those are web-traffic estimates, not user counts, and they leave out apps and API use. But the direction is clear. ChatGPT keeps growing in absolute terms while its share of the category shrinks.

So people are not abandoning ChatGPT. The problem for OpenAI is that its best users have become multi-tool users.

Is OpenAI still growing revenue as Anthropic pulls ahead?

Yes. OpenAI’s revenue is still growing very fast. But Anthropic is now ahead on reported revenue, which was hard to imagine a year ago.

OpenAI’s own curve is steep. It said its annualized revenue passed $20 billion at the end of 2025 and reached about $2 billion a month, or $24 billion a year, in March 2026. By September 2026, it told investors the pace was nearing $50 billion, according to the Financial Times and Reuters.

The $50 billion figure also corrected an earlier press figure of nearly $70 billion. The gap mostly comes from how revenue shared with cloud partners is counted. It is not a collapse in sales.

Recognized revenue tells a slower story. OpenAI reportedly booked $5.7 billion in Q1 2026 and $6.7 billion in Q2. The Financial Times reported a forecast of about $36 billion for the full year.

Anthropic grew faster. It reportedly booked $4.7 billion in Q1 2026, $11.6 billion in Q2 and $16.8 billion in Q3, up about 45% on Q2. On the same quarter, Q2, that is $11.6 billion for Anthropic against $6.7 billion for OpenAI.

One caveat matters. OpenAI reportedly counts partner sales net of the cloud provider’s share, while Anthropic counts some of them gross. That inflates Anthropic’s figure relative to OpenAI’s. So Anthropic is ahead on reported numbers, but the exact size of the lead is not settled.

OpenAI is also adding new revenue lines. Enterprise was more than 40% of revenue in March 2026, and ChatGPT ads reached a $1 billion run-rate in August.

So no, OpenAI is not dying on revenue. It doubled its pace in six months. It just stopped being the fastest-growing lab.

Is OpenAI’s compute bill the real danger?

Yes. This is still the strongest OpenAI bear case, and the numbers keep growing.

In January 2026, OpenAI’s CFO said the company ended 2025 with about 1.9 gigawatts of compute, up from 0.6 gigawatts in 2024 and 0.2 gigawatts in 2023. Since then, the commitments have kept piling up. In August 2026, Nvidia backed a campus in Ohio with up to 8 gigawatts planned for OpenAI under a 20-year lease. That sits on top of deals with Oracle, Microsoft, AWS, CoreWeave, AMD, Broadcom and Cerebras.

Those deals overlap, so they cannot simply be added up. No verified total of OpenAI’s committed gigawatts exists. But the Wall Street Journal reported in October 2026 that OpenAI was discussing roughly $600 billion to $750 billion of compute spending through 2030.

That changes how we should read the revenue growth. In normal software, more users usually mean better margins. In frontier AI, more usage can mean a bigger bill, especially when people move from chat to agents, coding and long automated tasks. OpenAI’s gross margin was about 39% in Q1 2026, according to The Information.

Execution is a risk too. On October 15, 2026, OpenAI said Sachin Katti, who led its compute buildout, would leave on November 15. The Financial Times also counted at least 47 major US data-center projects stalled or canceled in 2026, across the industry.

So if OpenAI ever gets into real trouble, it will not be because nobody wants it. It will be because it signed up for more compute than its revenue can pay for.

If you want more recent data on this point, see the latest figures in our Frontier AI Labs report ↓.

Is OpenAI running out of money?

No, not now. OpenAI has a large cash cushion and keeps raising more. The question is how long the cushion lasts.

OpenAI held about $73 billion in cash at the end of March 2026, according to The Information. It burned about $3.7 billion in cash in Q1. On October 1, 2026, Nvidia and SoftBank reportedly paid their final $10 billion installments in the $122 billion round, which closes that money in.

The losses look scary on paper. OpenAI’s Q1 2026 net loss was reportedly above $21 billion. But about $12.4 billion of that came from accounting on convertible securities, not cash spent. The operating loss was about $9.3 billion.

The long-term plan is what worries investors. The Financial Times reported a projected ~$280 billion of cumulative negative free cash flow through 2030, against about $350 billion of forecast revenue in 2030. OpenAI has not committed to a break-even year.

That is why OpenAI is reportedly seeking at least $30 billion more at around $1.4 trillion. That round has not closed, so $852 billion remains its last priced valuation.

So OpenAI is not running out of money. But its plan only works if investors keep funding it at very large scale until revenue catches up.

Has OpenAI lost the model race?

Not lost, but it no longer leads. In October 2026, OpenAI is one of four or five frontier labs, and it does not top any of the main rankings.

Look at the leaderboards on October 20, 2026. On Arena’s overall text ranking, Google’s Gemini 4 Argon was first, Meta’s Muse Wonder second and Anthropic’s Claude Opus 5.5 third. Google’s lead is preliminary and within the margin of error. On coding for the web and on agent tasks, Anthropic’s Opus 5.5 was first and OpenAI’s GPT-6 Astra second. Artificial Analysis put Anthropic first with 58 points, Meta at 56, and OpenAI and Google tied at 53.

Two of those models arrived in October. Google released Gemini 4 Argon broadly on October 13, 2026. Meta returned to the frontier with Muse Wonder on October 12 and released an open-weight model, Muse Spark Open, on October 15.

Price makes the picture harder for OpenAI. GPT-6 Astra lists at $10 per million input tokens and $50 per million output tokens. Claude Opus 5.5 lists at $4 and $20. Gemini 4 Argon launched at $2 and $10. DeepSeek’s cheapest current model costs $0.15 and $0.60 off-peak. OpenAI is selling a model that ranks second at the highest price.

Cheaper rivals are winning volume. On OpenRouter, a popular routing service for developers, DeepSeek had 24.4% of text requests in the week of October 12, 2026, ahead of Google at 19.7% and OpenAI at 18.5%. Chinese labs also keep raising money: Moonshot reached a private valuation of about $50 billion in October, according to Bloomberg. In Europe, Mistral released Large 4 and raised €3 billion at a valuation above €21 billion.

GPT-6 still matters. It reportedly cost Anthropic some ground with customers after its release. But the old idea that the best model always comes from OpenAI is over. The market now picks models by job and by price.

Is Anthropic winning the coding-agent race?

Yes. Anthropic still leads coding agents in October 2026, on both usage and revenue signals.

The best independent data comes from developers. In JetBrains’ survey of more than 15,000 professional developers, 39% used Claude Code at work in May–July 2026, up from 18% in January. In the US, it was 47%. Codex was at 16%, up from 3%. Among developers who named a main AI coding tool, 31% chose Claude Code.

Money points the same way. Claude Code passed a $2.5 billion annualized revenue pace in February 2026, the last figure Anthropic gave for it alone. Menlo Ventures estimated that Anthropic took 54% of enterprise spending on coding models in 2025, versus 21% for OpenAI.

Coding is also, by most accounts, the main reason Anthropic passed OpenAI on reported revenue. Barclays, for example, said on October 1, 2026, that it wants half its developers on Claude Code by the end of 2026.

That is why this matters so much for OpenAI. Consumer chat is OpenAI’s home field. Coding is the first place where AI became a clear budget line, and Anthropic got there first.

If you want more recent data on this point, see the latest figures in our Frontier AI Labs report ↓.

Is Codex catching up with Claude Code?

Codex is growing fast, but it has not caught up. It is closing the gap from far behind.

OpenAI’s own numbers are strong. On June 2, 2026, it said Codex had more than 5 million weekly active users. In June, Codex also generated 64% of the output tokens across Codex and ChatGPT inside enterprise accounts, according to OpenAI. That shows how much heavy work runs through it.

Since then, OpenAI has stopped giving Codex figures on their own. On July 21, 2026, it reported about 10 million weekly users across Codex and ChatGPT Work combined. On October 8, an OpenAI product lead said the two together had reached 40 million active users, without saying over what period. Those numbers cannot be compared with the earlier ones.

The bigger change is that Codex is now built into ChatGPT. On July 9, 2026, the Codex desktop app became the ChatGPT desktop app, and Codex is available on every plan, including Free. That gives Codex a distribution advantage Claude Code cannot match.

But adoption at work still favors Anthropic. JetBrains measured 16% workplace use for Codex against 39% for Claude Code, and OpenAI does not disclose Codex revenue. So we cannot compare the two products’ economics.

So Codex is not failing. It is one of OpenAI’s most important growth bets. But Claude Code still has the stronger brand with professional developers, and Codex is still chasing it.

Are enterprises moving away from OpenAI?

No. Enterprises are still adding OpenAI. What they are not doing is choosing OpenAI alone.

OpenAI’s business reach is wide. It said in September 2026 that it served more than 2.5 million businesses, and enterprise made up more than 40% of its revenue in March. AWS began offering OpenAI models, Codex and managed agents on Amazon Bedrock in April 2026. Dell brought Codex to on-premise deployments in May. On October 6, Atlassian said GPT-6 models would power its enterprise agents.

OpenAI also got more room to sell. On April 27, 2026, it and Microsoft amended their deal. Microsoft’s license became nonexclusive, and OpenAI can now serve its products across cloud providers.

The pressure is in share, not in absolute sales. Menlo Ventures estimated that Anthropic took 40% of enterprise spending on model APIs in 2025, OpenAI 27% and Google 21%. In 2023, OpenAI had 50%. No newer independent measure was found.

And almost nobody runs a single model. A Stacklok survey in September 2026 found that about 92% of large North American companies used more than one model provider, and 25% used five or more. Companies want leverage on price, backup options and control over data.

So OpenAI is not being rejected by enterprises. It is competing in a normal enterprise software market now, on security, cost and integration, against rivals that are often cheaper or better at coding.

If you want more recent data on this point, see the latest figures in our Frontier AI Labs report ↓.

Is OpenAI’s product strategy still messy?

Less than it was in June. OpenAI has turned its many products into one app. The risk now is concentration, not confusion.

In June 2026, reports said OpenAI was preparing the biggest ChatGPT overhaul since launch. It shipped on July 9, 2026. ChatGPT Work added an agent that can work for hours across a user’s files and apps, plus plugins for business tools, scheduled tasks and computer use. The Codex desktop app became the ChatGPT desktop app. OpenAI also began winding down its standalone Atlas browser.

GPT-6 followed. From October 8, 2026, it rolled out to every ChatGPT user, including the free tier.

At the same time, OpenAI kept cutting what was not central. The Sora web and app experiences closed on April 26, 2026, and the Sora API was scheduled to end on September 24, 2026.

The trade-off is that almost every bet now lives inside ChatGPT. If the work features do not convert free users into paying ones, OpenAI has fewer side products to fall back on.

So the strategy is no longer random. It is a forced consolidation around ChatGPT, agents, Codex and monetization.

Is OpenAI losing too many senior leaders?

Yes, the departures have become a real warning sign. They are not an exodus, but the list has grown long in 2026.

In April 2026, Kevin Weil and Bill Peebles left as OpenAI shut Sora and folded its science work. Srinivas Narayanan, CTO for business applications, left around the same time. In June, Barret Zoph, head of enterprise sales, left, and Clive Chan, an early engineer on OpenAI’s custom chip program, joined Anthropic.

The summer exits reached the top. On July 13, 2026, Fidji Simo, who ran OpenAI’s applications business, moved to a part-time advisory role for health reasons. Her duties were split among Greg Brockman, Sarah Friar and Jason Kwon. Johannes Heidecke, head of safety, also left in July. On August 11, Brad Lightcap, COO until April, said he would leave to start something new.

Then on October 15, 2026, OpenAI said Sachin Katti, who led its compute buildout, would leave on November 15. Uday Ruddarraju takes over the compute organization.

The pattern matters more than any single name. The exits touch products, enterprise sales, safety and compute, the areas where OpenAI must execute in 2027. Some exits are personal and successors are named, so this is not a fatal drain. But it is a risk investors will price.

If you want more recent data on this point, see the latest figures in our Frontier AI Labs report ↓.

Are lawsuits becoming a bigger OpenAI problem?

Yes. Legal risk has grown in 2026, and it is now led by states, not just publishers.

On June 1, 2026, Florida sued OpenAI and Sam Altman, saying ChatGPT was marketed while serious risks were hidden, including risks to children. On October 15, 2026, at least eight more states followed, including Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska and West Virginia. They claim ChatGPT broke consumer-protection laws by harming children, giving dangerous advice to adults and misleading the public about safety testing.

There was one win. In May 2026, a court dismissed Elon Musk’s claims against OpenAI on time-limit grounds, though Musk was reported to plan an appeal. The New York Times copyright case and other publisher suits are still running.

The state suits are a different kind of problem. Copyright cases raise costs. Consumer-protection cases can change the product itself: age checks, default safety settings, how ChatGPT responds to people in distress, and how it is marketed. They also land just as ChatGPT becomes an agent that acts for users, which raises the stakes.

So lawsuits will probably not kill OpenAI. But a coordinated state campaign can slow it, raise compliance costs and weigh on its IPO.

Are ChatGPT ads working, or hurting trust?

Both, so far. Ads are working financially faster than many expected. The trust question has not gone away.

On August 31, 2026, OpenAI said ChatGPT ads had reached a $1 billion annualized run-rate less than 200 days after launch. It said tens of thousands of advertisers were buying, in more than 40 countries.

That is a fast start, but still small. $1 billion is about 2% of OpenAI’s ~$50 billion revenue pace. An Emarketer analyst said the pace makes a reported $2.5 billion ad target for 2026 unlikely, and agency results have been mixed.

The risk is also clear. OpenAI says ads are labeled, do not affect answers and can be controlled by users. But people ask an assistant for judgment, about doctors, schools, software or travel. The line between an answer and an ad is more sensitive than in search. And the state lawsuits now put ChatGPT’s marketing under a microscope.

So ads are not a death signal. They show OpenAI turning into a large ad-funded platform. That can work, but it changes how users see the product.

If you want more recent data on this point, see the latest figures in our Frontier AI Labs report ↓.

Why did OpenAI push its IPO to 2027?

Because it can still raise private money, and because a 2026 listing would have come right after Anthropic’s. The delay buys time, but it also shows pressure.

OpenAI filed IPO paperwork confidentially with the SEC in June 2026. On September 29, reports said Sam Altman had ruled out a 2026 listing and pushed it to 2027. Instead, OpenAI is reportedly seeking at least $30 billion in private funding at a valuation of around $1.4 trillion. Those talks are not closed, so $852 billion is still its last priced valuation.

Anthropic is moving first. It filed confidentially in June too, and it is targeting a listing in mid-November 2026, according to reports. A valuation of around $2 trillion has been discussed, with some investors closer to $1.5 trillion. Its last completed round, in May 2026, valued it at $965 billion, above OpenAI.

Anthropic’s draft prospectus also shows what public investors will see. It reportedly lists more than $600 billion of noncancelable infrastructure commitments. OpenAI will face the same questions: revenue quality, margins, compute obligations, lawsuits and a path to profit.

Waiting lets OpenAI see how markets price Anthropic and show more ad and enterprise revenue first. The cost is more dependence on huge private rounds, while its rival gets public capital.

So OpenAI is not being forced into an IPO. It is being forced to keep raising money until it is ready for one.

Is ChatGPT still the default AI brand?

Yes. ChatGPT still owns the default AI brand for the mainstream market.

User behavior is the proof. More than 1.2 billion weekly users in October 2026 is in a different league from any rival app. Meta’s new Muse app, for example, had about 2.5 million daily users in mid-October, according to Sensor Tower. ChatGPT is still where most people meet AI.

The paid base matters too. More than 50 million consumer subscribers in March 2026 means tens of millions of people pay for ChatGPT, not just try it. Paid habit is much harder to dismiss than free curiosity.

But default status can hide weakness at the edges. ChatGPT’s share of chatbot web visits fell to about 57% in August 2026, by Similarweb’s estimate. The highest-value users are moving serious work into Claude Code, Gemini, Cursor, open models or internal tools. That is how platforms weaken slowly: the mainstream stays, but the people who pay the most stop treating it as the best place to work.

So, is OpenAI dying soon?

No. OpenAI is not dying soon.

The facts do not support that conclusion. OpenAI has more than 1.2 billion weekly ChatGPT users, revenue running near $50 billion a year by its own investor figures as reported, about $73 billion of cash at the end of March, a completed $122 billion round, a $1 billion ads business, GPT-6 in every ChatGPT account, and Codex built into its main app. That is not a death spiral.

But the easy version of OpenAI is gone.

That version was simple: best model, best chatbot, fastest growth and no serious challenger. In 2026, each of those claims broke. Anthropic passed OpenAI on reported revenue and leads coding agents. Google and Anthropic top the model rankings. Cheaper Chinese models win developer traffic. Senior leaders have left. At least eight more states have sued. And Anthropic, not OpenAI, is on track to face public investors first.

The company is still massive, still growing fast and still one of the most likely long-term winners in AI. But it is no longer protected by being first.

From here, OpenAI has to prove three things at once: ChatGPT can earn more per user through work tools and ads, Codex can catch Claude Code, and revenue can keep pace with hundreds of billions of dollars of compute.

That is a much harder game.

Question	Conclusion	Explanation
Are ChatGPT users leaving?	No	ChatGPT passed 1.2 billion weekly users in October 2026. Its share of chatbot web visits is falling as users add Claude and Gemini.
Is OpenAI still growing revenue?	Yes	Its annualized revenue neared $50 billion in September 2026, per reports. But Anthropic is ahead on reported quarterly revenue.
Is compute becoming the real risk?	Yes	Reported compute spending plans reach $600 billion or more through 2030. The danger is that costs grow faster than revenue.
Is OpenAI running out of money?	No	It held about $73 billion in March 2026 and is seeking more. But it is projected to burn cash through 2030.
Has OpenAI lost model leadership?	Partly	In October 2026, Google and Anthropic top the main leaderboards. GPT-6 is close behind, at a higher price.
Is Anthropic winning coding agents?	Yes	Claude Code leads Codex 39% to 16% in workplace use among developers surveyed by JetBrains, and has a disclosed revenue run-rate.
Is Codex catching up?	Not yet	Codex grew fast and now ships inside ChatGPT on every plan. It still trails Claude Code with professional developers.
Are enterprises leaving OpenAI?	No	AWS, Dell and Atlassian deals show expansion. But about 92% of large companies use several model providers.
Is OpenAI’s strategy too messy?	Less so	ChatGPT Work and GPT-6 merged chat, agents and Codex into one app. The new risk is that every bet depends on ChatGPT.
Is OpenAI losing key talent?	Yes	2026 exits include its applications chief, former COO, head of safety and compute chief. A warning sign, not a collapse.
Are lawsuits becoming more dangerous?	Yes	Florida sued in June 2026, then at least eight more states in October. That can change the product and slow it down.
Are ads a trust problem?	Maybe	Ads hit a $1 billion run-rate in August 2026, but they change how users see an assistant’s neutrality.
Why did OpenAI delay its IPO?	Time and money	It pushed its IPO to 2027 and seeks private money instead, while Anthropic targets a November 2026 listing.
Is ChatGPT still the default brand?	Yes	ChatGPT remains the mainstream AI default, with over a billion weekly users and tens of millions of paying subscribers.
Is OpenAI dying soon?	No	The evidence shows growing pressure, not death. OpenAI is not dying, but it is no longer the clear leader.