Whether AI is a bubble depends on which argument you weigh more: the bull case points to profitable giants funding real technology, while the bear case points to unprofitable AI firms, circular deals and an economy leaning on AI spending. Everybody is talking about an AI bubble. This breakdown lays out the strongest arguments on both sides, then looks at what could make a bubble pop and what ChatGPT itself said when asked.
It is not financial advice. The aim is to present both cases fairly, and both turn out to have compelling points.
Watch the full video here, or skip to the breakdown below.
The AI bubble debate at a glance
- The bull case: not a bubble, or not yet (0:41)
- The bear case: usefulness is not profit (6:38)
- Bubble versus recession (12:42)
- What does AI think? (16:19)
1. The bull case: not a bubble, or not yet
A common claim is that AI feels like the dot-com bubble all over again, with endless money thrown at a revolutionary technology that has not yet changed the world. Look closer, the bull case says, and the two are quite different.
In the late 1990s the internet was new and poorly understood, and investors poured money into anything with .com in its name, profitable or not. Pets.com sold pet food online, had no plan to make a profit and was valued in the hundreds of millions. When investors realized many such firms had no real foundation, thousands of companies went bankrupt almost overnight.
Today’s AI money mostly comes from some of the biggest, most established and most profitable companies in the world: Google, Microsoft, Amazon and Meta. Their quarterly reports show record profits despite hundreds of billions spent on AI. Microsoft is not going into debt for AI and Google is not mortgaging its future. They are reinvesting profits from businesses that already work, so the money is real.
The technology differs too. The late-90s internet was full of promise but not yet useful to most people, who could not stream video, store photos in the cloud or use maps on a phone. AI is useful today. Millions of people use it daily for writing, coding, images and automating parts of their jobs, saving time and sometimes money. It is also built on an existing base of cloud computing, global data centers, powerful GPUs and decades of software, so even if individual projects fail, the servers, chips and algorithms will not disappear.
Finally, the funders are trillion-dollar giants with the capital and talent to keep investing as long as it takes. A bubble pops when there is no real value underneath, and here something real is being built.
AI may not make everyone rich overnight, but for once the hype may be pointing at something that matters.
2. The bear case: usefulness is not profit
For each reason to call it a non-bubble, the bear case has an equally strong reply. First, many people are not finding AI useful. A study cited in the script found around 95 percent of companies that tried to integrate generative AI reported zero productivity improvement, and some got worse, while only about 5 percent saw a positive return. For every company that benefits, nineteen waste money forcing AI where it does not fit. Usefulness is not the same as profitability.
Second, there is a gap between what AI was supposed to be and what it is. Talk of artificial general intelligence, machines that think like humans, has so far delivered AI girlfriends, image generators and video tools. OpenAI’s Sora 2 makes realistic videos, but it is being used as a glorified meme generator. Google’s AI Mode search is better than old search, but not a hundred times better. If this is the technology that changes everything, why does it feel like a minor upgrade?
Third, and strongest, the companies leading the boom mostly are not making money. OpenAI, Anthropic and Google’s AI division are described as massively unprofitable. Every prompt takes heavy computing power and electricity, so each question costs OpenAI money, and paid plans do not cover it. Nvidia is the exception, because it sells the chips: the shovels in a gold rush. Builders bleed while suppliers boom, and the bleeding is kept alive by investment rather than profit.
That is where round-tripping comes in, with companies funding each other’s growth in a circle. Nvidia announced it would invest up to $100 billion in OpenAI, and OpenAI confirmed it would spend a large share on Nvidia chips, which Nvidia books as revenue. OpenAI also promised AMD a large chip purchase and, as part of the deal, gets 10 percent of AMD’s stock. Oracle signed a $30 billion cloud and data center deal with OpenAI, and that money flows back to Nvidia and AMD for chips.
On paper everyone is growing, but it is the same money passed in circles. If investment dried up, OpenAI would stop buying Nvidia chips, Nvidia would stop investing back, Oracle would stop building data centers and the whole market could come crashing down.
3. Bubble versus recession
If we are in a bubble, expect the word recession. But there is a difference between a bubble popping and causing a recession, and a recession popping the bubble. In 2008 the housing bubble grew so large that its burst took the economy down with it. In the other scenario, a slowing economy makes investors nervous and they stop funding high-risk sectors like AI. The script says current data suggests the second is the one we may be heading toward.
According to JP Morgan, about 30 companies heavily invested in AI make up around 44 percent of the S&P 500’s total value. Nearly half the market is concentrated in firms tied to the same technology. Reports also suggest the AI boom is hiding how weak the rest of the economy is: in 2025 around 40 percent of America’s real GDP growth came from tech companies spending money, mostly on AI. Without it, the US economy would have grown only 0.1 percent in the first half of 2025, and some analysts say it might already be in recession.
If the non-AI economy suffers too much and investors pull money out of AI to cover losses elsewhere, the biggest names are the most exposed: Microsoft, Google, Nvidia and Amazon. Smaller startups would shut down in large numbers, causing layoffs across the economy, and the industry would shrink to a handful of massive players able to keep going.
That is the irony. The companies that turned AI into a bubble may be the only ones big enough to survive its pop. And bubbles also reset things by clearing out the noise and forcing real value to show. Like the internet, if AI is the next great technological revolution it will survive the pop, though it may look very different on the other side.
A bubble that causes a recession and a recession that pops a bubble are two different events.
4. What does AI think?
To finish, the team gave ChatGPT the script for this video, asked it to analyze the arguments and requested a one-word answer to whether we are in an AI bubble. It thought for 12 seconds and answered: probably.
Neither side is declared the winner here, and nothing in the video tells you what to do with your money.
Where that leaves the debate
Both cases rest on real facts. The bull case has cash-rich funders, daily users and existing infrastructure, while the bear case has unprofitable labs, circular deals and a stock market and economy leaning heavily on AI spending.
Whatever happens, the video’s closing point is that AI is never going to replace your intelligence, curiosity and willingness to learn.
FAQ
Is AI a bubble like the dot-com bubble?
The bull case says no, because AI is funded by profitable giants like Google and Microsoft rather than unprofitable startups, and it is already useful. The bear case says leading AI firms lose money and rely on investment instead of profit.
What is round-tripping in AI?
Companies funding each other’s growth in a closed circle. For example, Nvidia invests in OpenAI, OpenAI spends on Nvidia chips, and Nvidia counts that as revenue.
Why do some say AI is not profitable?
Running large language models takes huge computing power and electricity, so each prompt costs money. OpenAI, Anthropic and Google’s AI division are described as unprofitable, while chip seller Nvidia profits.
How concentrated is the stock market in AI?
JP Morgan is cited as saying around 30 AI-heavy companies make up about 44 percent of the S&P 500’s total value.
Could a recession pop the AI bubble?
Yes. If the wider economy slows, investors may stop funding high-risk sectors like AI. The script notes about 40 percent of 2025 US real GDP growth came from tech spending, mostly AI-related.
What did ChatGPT say when asked if we are in an AI bubble?
Given the script and asked for one word, it thought for 12 seconds and said probably.
