OpenAI's AI Crisis Is Worse Than You Think

OpenAI's crisis is worse than you think, so let me give you three numbers. 218 billion dollars. That's how much money OpenAI expects to burn through 2029. Not revenue burned, gone.

295 percent. That's how many ChatGPT uninstalls jumped in a single day after OpenAI signed a deal with the Pentagon. And number one, that is where Claude, the rival app for Anthropic, landed on the U.S. App Store for the first time ever.

Whilst ChatGPT was failing on three numbers, three different problems, all here in the same company at the same time. And none of these things happened by accident. They were all connected. The money problem created the pressure that led to the Pentagon deal.

The Pentagon triggered the user revolt, and the user revolt handed Claude the number one spot. It's just a chain reaction, and it started with the money. Because this is not just OpenAI having a rough few weeks. This is the biggest financial gamble in the history of business.

A company burning more money than any startup ever, trying to outrun rivals who give the same thing away for free. Whilst a boycott movement called QuitGPT has 2.5 million people pledging to cancel their subscriptions. If you use AI tools in your business, you need to understand what is actually happening here. Not the headlines, the full story.

So let's talk about the money problem. ChatGPT hit 100 million users faster than any product in history. For two or four years, it said AI to someone on the street. They fought ChatGPT.

It was the default, right? So the real financial numbers feel like someone made a mistake in the spreadsheet. OpenAI generated around $13.1 billion revenue in 2025, which sounds like a rocket ship, until you see they posted a $13.5 billion net loss in just the first half of 2025. In six months, they lost more money than they made in the entire year.

OpenAI spends $1.69 of every dollar of revenue it generates, and the drain is getting bigger. Updated forecasts show cash burn of $25 billion in 2026 and $57 billion in 2027. Cumulative cash burn reaches $665 billion through 2030, with the company not expecting positive cash flow until that year. $665 billion before they make a single dollar of real profit.

That number is bigger than the entire company, sorry, the entire economy of most countries. Now, I get the skepticism. Amazon lost money for years. Netflix burned cash for a decade.

This is just what disruption looks like, right? But there's a critical difference. Uber lost billions before its IPO, but it was building a physical network of drivers in thousands of cities. Tesla struggled for years, but it was building factories, and the global charging network competitors could not copy overnight.

And both companies were building something real, something physical, something that got harder to compete with the bigger it got. OpenAI is building something that evaporates. No physical network, no physical assets. The primary product loses money every time someone asks it a complex question, and there's nothing stopping users from leaving tomorrow.

So that's why it costs as much. Now, AI does not scale like normal software. If you want a regular app to be twice as good, you hire better engineers. You spend roughly twice as much, and this cost scales in a straight line.

AI scales like a rocket that gets heavier the higher it goes. Training cheaply costs roughly $100 million for one training run. The models arriving right now, each training could run over $1 billion. One attempt to teach a model, and you can't do it twice and walk away.

And the competition releases something better, and then your model gets stale, you run it again, each time the bill gets bigger. OpenAI is locked in a cycle, where they must keep spending impossible amounts, just to stay one step ahead of companies giving away the same technology for free. And then there's the hardware problem. Each NVIDIA Blackwell B200 chip costs $30,000 to $40,000.

You need tens of thousands wired together. And unlike a factory machine that runs for 20 years, these chips have a shelf life measured in months. Imagine a trucking company that has to buy an entirely new fleet every 18 months. Not because the old ones broke, because the roads changed, then add the electricity.

Project Stargate, their $500 billion supercomputer gamble, demands the equivalent of multiple nuclear reactors. OpenAI is now negotiating directly with nuclear power plants and solar farms, just to keep the lights on. And every free chat GPT user makes this worse. Hundreds of millions of people using the free tier.

Every session costs OpenAI real money. The product works, the bills are just impossible. And when people hear these numbers, the question is always the same. How are they still open?

Microsoft. But not the way you think. When Microsoft invests billions into OpenAI, a massive chunk never actually leaves Microsoft. Instead of cash, Microsoft gives OpenAI cloud credits, like a gift card that can only be spent at the Azure store.

OpenAI records it as money raised. It looks like capital in the books. But those credits get spent at Azure, recycling the investment back into Microsoft revenue stream. And the dangerous part here is that you can't pay employees with gift cards.

When OpenAI hires a top AI researcher at $2 million a year, they need actual dollars. Every quarter, OpenAI has to raise hard cash from outside investors just to cover what the Microsoft credits cannot touch. And the circular financing runs deeper. NVIDIA committed up to $100 billion to OpenAI, money that, as OpenAI's own CEO acknowledged, goes right back to NVIDIA in GPU purchases.

NVIDIA also invests in CoreWeave, which supplies cloud capacity to OpenAI and has spent billions buying NVIDIA chips. So NVIDIA invests in OpenAI. OpenAI buys NVIDIA chips. NVIDIA uses that revenue to invest more.

The money is going in circles. And the real question is whether OpenAI can break out the loop before the loop breaks them. Even if OpenAI solve the hardware and electricity problems tomorrow, they face a three-front war that is getting harder to win. Meta released their Lama models for free, not charity.

A tactical strike. When Zuckerberg gives everyone top-of-the-line AI for free, he sets a ceiling on what OpenAI can charge. And Meta's real business is ads on Instagram and Facebook. They don't need AI to be profitable.

They just need to make sure nobody else can charge a lot for it. OpenAI's only product is the AI. They have nothing else. Two, Chinese competitors.

Models now priced at one-sixth the cost of comparable US systems. DeepSea launched V4, a one-trillion-parameter coding model, in February 2026, a frontier capability a fraction of the cost. And it breaks the central assumption OpenAI's business rests on, that you need to spend tens of billions to stay at the frontier. If that assumption is wrong, your justification for burning $218 billion collapses.

And three, the user's chat cheapities, web traffic, share fell from 86% in January 2025 to 64% in January 2026. Gemini grew from 5% to 21%. About 75% of OpenAI's revenue comes from consumer subscriptions, but cancellations are rising as well. And only 20% to 30% of enterprise users are sticking with OpenAI's API long-term.

Users are mercenaries. Someone offers something close enough for cheaper, they leave. They download a different app in 30 seconds. There's no lock-in, no switching cost.

And then came February 2028. So on February the 27th, 2026, the Department of Defense approached both OpenAI and Anthropic. They said, give us unrestricted access to your AI systems for military use. Anthropic's CEO, Dario Amodi, said no.

He wrote publicly that he cannot in good conscience accede to the Pentagon's request. Some uses, he said, are simply outside the bounds of what today's technology can safely do. Within hours, Sam Waltman signed the deal and the internet exploded. By the end of that week, quick chat GPT had over 2.5 million people pledging to cancel their subscriptions.

On Saturday, March the 1st, chat GPT mobile app uninstalls jumped to 295% compared to the day before. One-star reviews surged 775% in a single day. OpenAI reportedly lost approximately 1.5 million paid subscribers in the first week. At $20 per month, that's $30 million in lost monthly recurring revenue.

Gone from one week of the boycott. So where did they go? Downloads of Claude jumped 51% in a single day. By Sunday, Claude hit number one on the US App Store.

That had never happened before. Not once in the history of these two apps competing. Anthropic confirmed daily active users have more than tripled since the beginning of 2026 and paid subscribers doubled. So more than 1 million new signups per day.

Now, let me be fair. Chat GPT still has over 300 million weekly active users. One boycott week does not raise that, doesn't change that. But what it shows is this.

The loyalty that once made chat GPT untouchable is gone. Users will leave. They are leaving. The first time Claude is good enough that people are not just switching off protests.

They are switching and staying. That is a different world than one that OpenAI built in its business plan around. And let's talk about that. The AGI bet.

Underneath all the losses, the controversy, the competition, there is a single thesis. The bet is AGI. Artificial General Intelligence. An AI that can do any intellectual task a human could do.

But faster and cheaper. Run entire departments. Handle your taxes. Write your code.

Do the work of 50 people for the cost of a monthly subscription. If OpenAI builds that, the cash burn does not matter. Revenue would be effectively unlimited. But here's the brutal math.

If they're losing $14-17 billion a year, every month of delay costs over a billion dollars. If the breakthrough takes five years instead of two, they face a funding gap of nearly $100 billion. Nobody can tell you with certainty when AGI arrives. Or if it arrives with this approach at all.

HSBC analysis concluded OpenAI likely will not make money by 2030 and will still face a $207 billion funding shortfall. OpenAI could burn through $14 billion in 2026, risking bankruptcy by mid-2027. And that's only 15 months from now. The most likely outcome is not a dramatic crash.

It's something quieter. By mid-2027, cash reserves from the 2025 rounds will be nearly empty. OpenAI faces a choice. Raise another round at a lower valuation, crushing employee stock options.

Or just sell. Microsoft is the natural buyer. Over $80 billion in cash reserves. One of the few entities on earth that could sustain the burn rate.

The startup had started the entire AI revolution could get folded into a corporation. Not because it failed, but because it ran out of time. And this is the kind of thing we break down inside the AI profit boardroom every single week. Not just, here's a new tool, go try it.

We cover what's actually happening in the AI industry. What it means for your business. And what you should be doing right now to stay ahead. Four weekly AI automation coaching calls.

You get tutorials. There's 2,600 business owners in there. So there's always someone online who has solved the same problem you are stuck on. Link in the comments description or go to the AIprofitboardroom.com to get started.

Now let's talk about the talent war as well. Because everything I've described gets dramatically worse when you factor in the people inside OpenAI. Engineers at OpenAI are eight times more likely to leave for Anthropic than the reverse. CTO Mira Murati and co-founder Ia Sveka are among a wave of senior departures.

Many leaving to start competitors or join Anthropic. To stop the bleeding, OpenAI expects to spend $6 billion on stock-based compensation and nearly half of projected revenue on it. Not on research, not on infrastructure. On convincing their own employees not to quit.

And those retention packages only work as long as valuation keeps climbing. The moment investors stop believing the story, those stock packages are worth less. The researchers start doing the math and the math points to the door. Bring it all together.

So $218 billion in burn. 295% uninstalled surge. Claude number one on the app store for the first time ever. The money problem created the desperation that led to the Pentagon deal.

And the Pentagon deal triggered the biggest user revolt in AI history. The user revolt handed Claude the number one spot in Anthropic. One million new signups a day. Underneath all of it, a bet on AGI has to pay off within a very narrow window.

A window that gets smaller every single month. And I'm not saying that OpenAI is finished. They have hundreds of millions of users, billions in the bank, and the most recognizable brand in AI. They could actually pull this off.

They could get through it. But the AI landscape is not the same as it was six months ago. The tools are better. The competition is fiercer.

The users more informed. The companies are under more pressure. And the default choices are being questioned. So loyalties are shifting because of that.

Look at all the new updates that Claude has released recently. And here's the part of the story I find genuinely significant. Anthropic gained market share by declining business. They said no to a $200 million government contract.

And in return, they got one million new signups a day. Paid subscribers doubled. Number one on the app store. A company said no.

And by saying no, they actually won. And that tells you something important about where AI users are heading. Because they care about building these tools. They care about what those tools are used for.

They are paying attention in a way that they probably weren't 18 months ago. The people and the businesses paying close attention right now are going to be in a very different position than the ones just waiting to see what happens. And I wouldn't wait if I was you. I would pay attention.

I would learn the tools now. Understand the landscape right now. Because it's always changing, right? The new developments that come out all the time are just becoming better and better.

Because by the time it's obvious what happened, it will already be too late to get ahead of it. Now, if you want to go deeper on AI automation and actually start building this into your business, come find us in the AI Profit Boardroom. Link is in the description. And I'll see you in the next one.

Cheers.