Will OpenAI die soon?

Why do people suddenly think OpenAI is dying right now?

Because it looks like OpenAI no longer feels inevitable. It used to be huge and the only thing people were talking about. That is not the case anymore.

That is the tension. In the last few months, OpenAI has shown some of the strongest operating numbers in tech: 900 million weekly ChatGPT users in February 2026, 50 million consumer subscribers, roughly $2 billion in monthly revenue by March, and a $122 billion funding round at an $852 billion valuation. On those numbers alone, “OpenAI is dying” sounds absurd.

But the recent signals around the company are not all clean. Anthropic has turned Claude Code into a massive enterprise wedge.

ChatGPT is still growing, but Claude and Gemini are growing inside important user segments.

OpenAI is pushing Codex hard because coding agents are becoming the most monetizable AI workflow.

The company is testing ads, preparing a major ChatGPT redesign, discontinuing Sora, folding side projects, watching senior people leave, fighting a new Florida safety lawsuit, and moving closer to public markets.

So the real question we will try to answer is: over the last 6 to 9 months, do the real-world signals show a company compounding, or a company starting to lose control of the market it created?

Are people abandoning ChatGPT these days?

No. Recently, there is no serious sign that people are abandoning ChatGPT.

The freshest consumer signal is, actually, still brutally strong. In February 2026, OpenAI said ChatGPT had more than 900 million weekly active users and more than 50 million consumer subscribers. Then in early June 2026, Reuters reported Sensor Tower estimates that the ChatGPT app had crossed 1 billion monthly active users in May, making it the fastest app ever to reach that scale.

That does not look like abandonment. Actually, ChatGPT is still the default AI product for normal users.

But we should not stop there, because the second layer is more interesting.

Sensor Tower also estimated that Claude had 56 million monthly active app users in Q2 2026 and was growing 640% year over year, versus 62% for ChatGPT. Claude is still tiny compared with ChatGPT by raw app scale, but the growth difference tells us where the pressure is coming from.

Anthropic is not beating OpenAI in mass adoption, but it is definitely gaining faster in the segments where people actively compare AI tools.

That is the important interpretation. ChatGPT is being surrounded. A lot of serious users now keep ChatGPT, Claude, Gemini, Perplexity, DeepSeek, and coding agents in the same workflow. That weakens the old idea that ChatGPT is the only AI interface people need.

So, clearly, people are not abandoning ChatGPT these days. The problem for OpenAI is that the best users are becoming multi-tool users.

Is OpenAI still growing revenue right now?

Yes. Revenue is absolutely not the death signal.

In March 2026, OpenAI said it was generating $2 billion in revenue per month, up from $1 billion per quarter at the end of 2024. That means the company moved from roughly $4 billion annualized to roughly $24 billion annualized in a little over a year. There are very few software companies in history with that kind of revenue acceleration.

The composition is also getting more serious. Subscriptions are still huge, but OpenAI is increasingly monetizing through enterprise seats, API usage, Codex, business products, cloud distribution, and ads testing. Recent reporting also says business customers are becoming a much larger part of revenue, which matters because enterprise revenue is where OpenAI can justify higher prices and deeper workflow integration.

The March 2026 funding round says the same thing from another angle. OpenAI raised $122 billion in committed capital at an $852 billion post-money valuation. Investors do not give that kind of money to a company with weak demand. They give it to a company whose demand is so large that capital becomes part of the product strategy.

So no, OpenAI is not dying on revenue.

Is OpenAI’s compute bill the real danger these days?

Yes. This is probably the strongest OpenAI bear case right now.

OpenAI’s demand is massive, but serving that demand is not cheap software distribution. It is chips, energy, data centers, model training, inference capacity, reliability engineering, and cloud partnerships.

In January 2026, OpenAI’s CFO said the company ended 2025 with about 1.9 gigawatts of compute footprint, up from 0.6 gigawatts in 2024 and 0.2 gigawatts in 2023. That is almost a tenfold increase in two years.

That number changes how we should read OpenAI’s growth. In normal SaaS, more users often mean cleaner margins over time. In frontier AI, more usage can mean a bigger infrastructure bill, especially when users move from asking questions to generating images, running agents, writing code, processing files, and automating long workflows.

The last few months show OpenAI trying to solve exactly that. It raised a record private round in March. It expanded access through AWS. It is pushing Codex into enterprise workflows where the value per query is higher. It is testing ads for free and Go users. It is moving toward an IPO path. Those are not random moves. They all point to the same problem: OpenAI needs more monetization per unit of compute.

So if OpenAI ever gets into real trouble, it will be because everyone uses it too much, in ways that are still expensive to serve.

If you want more recent data on this point, please see our latest generative AI market report.

Is OpenAI losing the model race right now?

No, OpenAI has not lost the model race. But recently, the model race has stopped being one race.

That is the big shift. Two years ago, the question was mostly “who has the best chatbot?” Today, users split the answer by job. Claude can feel better for coding and long-form work. Gemini benefits from Google’s distribution and search context. Open models are good enough for cost-sensitive workloads. Specialized agents can beat general chatbots in narrow workflows.

This is why OpenAI’s position feels less dominant, even when the company is still technically frontier-class. The market is no longer asking one model to do everything, but many models to do specific jobs well: code generation, data analysis, enterprise search, spreadsheet work, voice, image generation, customer support, security, legal drafting, and personal assistance.

Recent user research points in that direction too. A 2026 cross-platform study of active AI users found that more than 80% used two or more AI platforms, and that Claude, ChatGPT, and DeepSeek had statistically similar satisfaction ratings. That is not a perfect market-share study, but it captures the user behavior that matters: engaged users do not behave as if one model owns the category.

So OpenAI is not technically dead. It is still one of the frontier labs. But the idea that model quality alone will protect it is much weaker than it was.

Did Anthropic steal the coding-agent story recently?

Yes. Anthropic has taken the sharper coding-agent story right now.

The recent Claude Code numbers are hard to ignore. In March 2026, Anthropic said Claude Code had passed $2.5 billion in run-rate revenue, more than doubled since January 1, and doubled weekly active users since the start of the year. It also said more than 500 customers were spending over $1 million annually across Claude.

It goes beyond “developers like Claude.” That is actually a concrete enterprise revenue signal. Coding agents are one of the first AI categories where companies can see ROI quickly: code migration, bug fixing, test creation, internal tools, documentation, pull requests, and security review. If Anthropic wins that workflow, it wins a budget line, not just attention.

What makes this especially dangerous for OpenAI is that Claude Code gives Anthropic a very clean story. Claude is not only another assistant, but it is now where teams build software. That is simple enough for developers, CTOs, and CFOs to understand.

OpenAI is not standing still, but Anthropic forced the fight onto better terrain. Consumer chat scale is OpenAI’s home field. Coding agents are more contested. And right now, Anthropic has the cleaner narrative there.

If you want more recent data on this point, please see our latest generative AI market report.

Is Codex actually failing these days?

No. Codex is not failing. It is growing fast, but it is still trying to catch Claude Code’s mindshare.

OpenAI’s recent Codex numbers are strong. On June 2, 2026, OpenAI said Codex had more than 5 million weekly active users, up more than 6x since the desktop app launched in February. It also said knowledge workers now represent about 20% of Codex users and are growing more than three times as fast as developers.

That last detail matters. OpenAI is not only trying to win coding but to turn Codex into a broader work-execution layer. Reports, spreadsheets, presentations, contracts, internal apps, data analysis, sales workflows, research work, and operational tasks are all moving into the Codex story.

The June 2026 product update makes that clear. OpenAI added Codex Sites, annotations, and role-specific plugins connecting business tools. In other words, Codex is being pushed beyond “AI pair programmer” and toward “agentic work platform.”

So, clearly, Codex is not dead. It is one of OpenAI’s most important current growth bets.

But the harder take is also true: OpenAI is now reacting to a market where Anthropic made the coding-agent wedge feel obvious first. Codex has the distribution advantage through ChatGPT. Claude Code has the sharper specialist brand. That is the fight.

Are enterprises moving away from OpenAI right now?

No, OpenAI is safe for now. The recent evidence points more to enterprise expansion than enterprise abandonment.

In April 2026, AWS announced that Amazon Bedrock would offer OpenAI models, Codex, and OpenAI-powered Managed Agents. In June 2026, OpenAI expanded that positioning further, making OpenAI capabilities easier to deploy inside AWS environments with existing enterprise controls. That is not what happens when enterprises are walking away.

The Dell partnership points in the same direction. In May 2026, OpenAI and Dell announced a path to bring Codex into hybrid and on-prem enterprise environments. That matters for regulated buyers that cannot easily send proprietary code or sensitive workflows into a normal public-cloud setup. It expands OpenAI’s reach into companies that were structurally harder to serve before.

But there is a more subtle pressure underneath. Enterprises are not becoming “OpenAI-only” shops. On the contrary, they are trying to take the multi-model pathway. CIOs want optionality across OpenAI, Anthropic, Google, AWS-native tools, open models, and internal systems. They want governance, pricing leverage, region control, data protections, and fallback options.

So OpenAI is not being rejected by enterprises. Instead, it is being forced into a more mature enterprise software market, where every deployment has to compete on security, cost, integration, and workflow value.

That is still a huge opportunity. It is just less magical than “everyone will use the same chatbot.”

If you want more recent data on this point, please see our latest generative AI market report.

Is OpenAI’s product strategy getting messy right now?

Yes, we have to be honest about that. Recently, OpenAI looks like a company trying to compress five strategies into one product.

In early June 2026, the Financial Times reported that OpenAI was preparing the biggest ChatGPT overhaul since launch, turning it into more of a superapp with Codex, agents, and third-party services. That is a big strategic clue. OpenAI does not want ChatGPT to remain a box where users type questions. The goal is a place where users execute work.

That explains several recent moves. Codex is being pushed into business workflows. Ads are being tested for Free and Go users. Enterprise distribution is expanding through AWS and Dell. OpenAI has filed confidential IPO paperwork. The product is being reshaped around higher-value usage.

At the same time, OpenAI is pruning what looks less central. Sora is the best example. OpenAI’s help center says the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API will be discontinued on September 24, 2026. Reports around the shutdown also said Disney’s planned OpenAI partnership was derailed after the video-app retreat.

That is not a small detail. Sora was one of OpenAI’s most visible wow-factor products. Shutting it down tells us something about the current phase of the company: flashy consumer AI is less attractive when the compute bill is high and the IPO story needs discipline.

So yes, the strategy looks messy, but not random messy. It looks like a forced consolidation around enterprise, agents, Codex, monetization, and public-market readiness.

Is OpenAI losing key people recently?

Yes, recent departures are a real warning signal. But the pattern is more specific than “everyone is leaving OpenAI.”

The old examples are not the point anymore. Mira Murati and Ilya Sutskever matter historically, but they are not the recent signal.

On April 17, 2026, TechCrunch reported that Kevin Weil and Bill Peebles were leaving OpenAI as the company shut down Sora and folded science work. Weil had moved from chief product officer into OpenAI for Science. Peebles was closely tied to Sora. Around the same period, Business Insider reported that Srinivas Narayanan, OpenAI’s CTO for B2B applications, was also leaving.

Then in June 2026, Clive Chan, one of the earliest engineers on OpenAI’s custom chip program, left for Anthropic. That one matters because the OpenAI story is increasingly about infrastructure. Losing a chip-program veteran to the company’s main frontier-lab rival is not just ordinary startup churn.

What makes this cluster important is the mapping. The departures touch science, video, B2B applications, and custom chips.

Those are exactly the areas where OpenAI is under current pressure: side projects being folded, enterprise execution becoming more important, and infrastructure becoming strategic.

Now, OpenAI does not have a fatal talent exodus. However, recent departures line up too neatly with OpenAI’s strategic reset to ignore.

If you want more recent data on this point, please see our latest generative AI market report.

Are lawsuits becoming a more immediate OpenAI problem?

Yes. Legal risk has become more immediate because it is no longer only about copyright.

The copyright fight still matters, especially the New York Times case and broader questions around training data. But the fresher issue is product safety. On June 1, 2026, Florida filed a civil lawsuit against OpenAI and Sam Altman, alleging that ChatGPT was marketed while serious risks were concealed, including risks to children and harmful behavior.

That is a different kind of legal problem. Copyright cases can raise licensing costs and constrain training practices. Product-safety cases can affect the interface itself: age gates, parental controls, memory, mental-health responses, safety defaults, disclaimers, and how aggressively ChatGPT can be marketed to the public.

The timing is also bad for OpenAI. This is happening while the company is testing ads, moving toward a possible IPO, and trying to make ChatGPT more agentic. A chatbot that only answers questions already raises safety questions. An agent that books, buys, writes, edits, schedules, and executes tasks raises more.

So legal risk probably does not kill OpenAI soon. But it can slow the company, increase compliance costs, and make the product less free to move fast.

Are ads a smart move or a trust problem for OpenAI?

Both. Ads are financially logical, but strategically dangerous for OpenAI.

OpenAI began testing ads in ChatGPT for Free and Go users in 2026. The company says ads are clearly labeled, answers remain independent, conversations stay private, and users keep control. That matters because OpenAI knows exactly what the trust issue is.

The financial logic is obvious. ChatGPT has enormous free usage, and free usage costs money. If OpenAI can monetize even a small share of free sessions through ads, it creates a new revenue line that does not depend only on subscriptions or enterprise contracts.

But AI ads are not like search ads. In search, users expect commercial results. In an assistant, users often expect judgment. If the product helps someone choose a doctor, a trip, software, a therapist, a school, or a business vendor, the line between answer and influence becomes more sensitive.

So ads are not a death signal yet. OpenAI is moving from “beloved magical tool” to “large monetized platform.” That can work. But it changes how users feel about the product.

If you want more recent data on this point, please see our latest generative AI market report.

Is OpenAI being forced toward an IPO faster than it wants?

Yes, probably. The recent IPO moves look like both opportunity and pressure.

On June 8, 2026, AP and other outlets reported that OpenAI had confidentially filed IPO paperwork with the SEC. OpenAI said it had no fixed timeline, but the signal is still important. The company is preparing access to public markets while AI infrastructure spending is exploding.

That matters because frontier AI is becoming a capital markets race. OpenAI, Anthropic, xAI, SpaceX-linked AI infrastructure, Google, Amazon, Microsoft, Meta, and Nvidia are all connected to the same question: who can fund enough compute, talent, and distribution to stay at the frontier?

OpenAI’s March 2026 private round bought time. An IPO would buy a different kind of currency: public capital, liquid shares for employees, acquisition flexibility, and a way to finance infrastructure at a scale private markets may not always absorb.

But IPO readiness also makes the company more exposed. Public investors will want to see revenue quality, margins, customer concentration, compute commitments, legal risks, and the path to profitability. That is a very different environment from the private-market story of “trust us, AGI is coming.”

Is OpenAI still the default AI brand today?

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

The strongest proof is still user behavior. A product estimated at 1 billion monthly active app users in May 2026, with 900 million weekly users disclosed by OpenAI in February, is not a niche product. ChatGPT has become the default consumer interface for AI.

The subscriber number also matters. More than 50 million consumer subscribers means people are not only trying ChatGPT. Tens of millions are paying for it. That is a major anti-death signal because paid habit is much harder to dismiss than free curiosity.

But default status can hide weakness at the edges. Casual users may stay with ChatGPT while the highest-value users move serious work into Claude Code, Gemini, Cursor, Perplexity, open models, or internal enterprise tools. That is how platforms weaken slowly: the mainstream still uses them, but the frontier users stop treating them as the best place to work.

So, is OpenAI dying soon?

No. As of today, OpenAI is not dying soon.

The recent facts do not support that conclusion. In the last few months, OpenAI disclosed 900 million weekly ChatGPT users, more than 50 million consumer subscribers, roughly $2 billion in monthly revenue, a $122 billion private funding round, Codex growth to more than 5 million weekly active users, AWS distribution, Dell enterprise deployment, ads testing, and confidential IPO paperwork. That is not a death spiral.

But the easy version of OpenAI is dying.

That version was simple: best model, best chatbot, strongest brand, fastest growth, and no serious challenger. The recent evidence says that world is over. Anthropic has a real coding-agent revenue story. Claude is growing fast from a smaller base. Codex is accelerating because OpenAI needs a stronger work-execution layer. Sora has been discontinued. Science work was folded. Senior people tied to side projects, enterprise apps, and custom chips have left. Legal risk is moving from copyright into product safety. And the company now has to explain itself to public-market investors.

The company is still massive, still growing, still deeply important, and still one of the most likely winners in AI. But it is no longer protected by inevitability.

From here, OpenAI has to prove three things at the same time: ChatGPT can become a real work platform, Codex can fight Claude Code, and revenue can outrun compute.

That is a much harder game.

Question	Conclusion	Explanation
Are ChatGPT users leaving?	No	Recent user data points to continued massive scale, not abandonment. The real shift is that serious users increasingly add Claude, Gemini, and other tools next to ChatGPT.
Is OpenAI still growing revenue?	Yes	OpenAI reached roughly $2 billion in monthly revenue by March 2026. Demand is clearly not the problem.
Is compute becoming the real risk?	Yes	OpenAI’s compute footprint has scaled almost 10x in two years. The danger is that usage grows faster than margins.
Has OpenAI lost model leadership?	Not really	OpenAI remains frontier-class, but the market now picks models by task. That weakens the old “one best model wins” story.
Did Anthropic steal coding momentum?	Yes	Claude Code has become the clearest enterprise coding-agent story, with strong 2026 revenue and usage signals.
Is Codex failing against Claude Code?	No	Codex is growing fast and passed 5 million weekly users in June 2026. The issue is perception, not adoption collapse.
Are enterprises leaving OpenAI?	No	AWS and Dell partnerships suggest enterprise expansion. The pressure is that companies increasingly want multi-model stacks.
Is OpenAI’s strategy too messy?	Yes	Recent moves suggest forced consolidation: ChatGPT superapp, Codex, ads, enterprise, IPO prep, and Sora shutdown all at once.
Is OpenAI losing key talent?	Yes	Recent 2026 exits touched science, Sora, B2B apps, and custom chips. That looks like a warning signal, not a collapse.
Are lawsuits becoming more dangerous?	Yes	The newer risk is product-safety litigation, not only copyright. That can affect ChatGPT’s interface and deployment speed.
Are ads a trust problem?	Maybe	Ads can monetize huge free usage, but they change how users perceive an assistant’s neutrality.
Is OpenAI rushing toward IPO?	Maybe	The confidential S-1 suggests OpenAI is preparing for public markets because the AI race is capital-intensive.
Is ChatGPT still the default brand?	Yes	ChatGPT remains the mainstream AI default, with massive user scale and tens of millions of paying subscribers.
Is OpenAI dying soon?	No	The recent evidence shows pressure, not death. OpenAI is not dying, but its old inevitability narrative is gone.

Who is the author of this content?

NEW MARKET PITCH TEAM

We track new markets so founders and investors can move faster

We build living "market pitch" documents for emerging markets: AI, synthetic biology, new proteins, and more. Instead of outdated PDFs or hallucinated LLM answers, our clients get a clean, visual, always-updated view of what's really happening: key players, deals, regulations, and signals that matter. Learn more about us.