What’s Trending: The AI Bubble Burst
August 26, 2025
Share this post:
Tuesday, August 19th, started with some shocking news from Meta: a restructuring of its AI division. This sent the markets into a frenzy, with people predicting an AI bubble crash. Is this something that might happen at all? In the near future?
Speculation is at an all-time high regarding AI, but this goes in more ways than one. Read all about this hot topic and what experts have to say about it in this short article.
Theron Mohamed wrote a piece for Business Insider exploring the potential bursting of the AI bubble. For this, he interviewed Eric Gordon, a University of Michigan professor specializing in financial markets and tech. In this article, they go over why they think it’s coming any day, what could happen, and the main reasons behind it.
Gordon first says that this market resembles the dot-com that crashed in the early 2000s. Other experts, like Kevin O’Leary, say that it’s nothing to worry about. So what was the dot-com boom, you might wonder. The dotcom burst happened in the early 2000s, when stockbrokers began investing in companies that had “.com” in their name, trying to capitalize on a new market. The problem was that most companies that had that ending had nothing to do with the internet, and thus led to overvaluation. When the market corrected, many workers lost their jobs, stocks became worthless, and thus there was a historic market crash.
“More investors will suffer than suffered in the dot-com crash, and their suffering will be more painful.”
In 2022, Gordon noted that more people have been investing in AI and/or tech companies, which make up a large part of the US stock market. Not only that, but pensions and retirement funds are some of the most commonly invested portfolios, which makes Gordon think that there’s going to be “more bowls of spaghetti,” meaning people having to cut costs wherever they can.
You’d think that tech media would also join the panic, but according to TechHQ’s Joe Green, they’re not even shocked by this turn of events. Apparently, this side of the media has always known that AI is a bubble. For starters, the math shows that this business is running at a loss.
AI isn’t profitable, he argues, because most people aren’t willing to pay for it. Going by Reuters numbers, there are 1.5 billion active users of ChatGPT, but only 15.5 paying users. This tallies to only a .96% conversion rate, that some respectable organizations doubt. According to Sensor Tower and Similarweb, OpenAI only has around 600 million active users per month, bringing up the conversion to 2.58%. Still not good, especially considering that they’re outpacing the competition.
People just aren’t paying for AI, even if companies keep touting it as an essential tool for business development. Green comes to the conclusion that AI isn’t as crucial as they want us to believe. People have been writing emails, essays, and presentations for many years; there is only so much to gain by writing them faster. In fact, he questions whether people are willingly using AI, as Microsoft’s Copilot and Google’s Gemini are pre-baked into their default programs, meaning that people have no choice but to use “smart” searches.
So, is the AI bubble bursting? When will it happen?
Well, maybe. The problem is clear: there is no clarity regarding AI companies. Their financials are muddy, and they are not cheap to run, with $3 billion as the commonly cited price tag on training a model. Which can be funded by paid subscriptions if they raise prices, but since people aren’t paying for them, there’s the issue of fewer queries, and thus less data to analyze, and so less accurate answers.
“The more users an AI platform has, the greater it costs the company running it, and the ratio of cost to revenue remains constant at any scale: every answer to a query is formulated on-the-fly.”
Stephen Diehl, a London-based software engineer and tech writer, has written extensively about the AI bubble popping he predicts will happen soon. In his blog, he has published a short opinion piece about how unsustainable the current demand for the development of large language models (LLMs) has harmed software engineering.
Unlike Green, he doesn’t question the usefulness of AI for day-to-day tasks; he acknowledges that it is pretty useful, whether for engineers, marketers, or whoever else can save time by using it. The problem lies in the fact that it’s been sold as a limitless tool when researchers believe that it’s currently close to hitting its limit. To prove his point, he notes the minimal differences between ChatGPT-3 and ChatGPT-4.
But there is a bigger problem, which goes back to the US market crash that Eric Gordon predicts: American companies are jumping onto this as if it were the gold rush. They keep investing billions and billions of dollars, expecting returns that, to this day, haven’t come, all the while China’s DeepSeek started running a more advanced model at a quarter of the price.
“The technology, which is genuinely limitedly useful, would finally be freed from the impossible burden of returning impossible returns. It could simply be a tool, and we could get back to the real, unglamorous, but ultimately more rewarding work of using it to build better things.”
AI is a genuinely useful tool that has made efficiency really cheap for many industries, but experts believe that it has been overvalued, and when the bubble bursts, the market will overcorrect itself. A crash is imminent, but specialists believe that it will happen like the dotcom, in which the tool gives way to a more prominent use, but not necessarily the speculation that it’s being used for currently.
WHAT DO YOU NEED TO FIND?
