Revisiting the AI bubble theory: more cause for concern?

Even OpenAI CEO Sam Altman now admits that investors might be a little overexcited about AI. He says he’s not worried because it’s only natural – but plenty of others are. In Silicon Valley, fears over the AI bubble bursting are growing: it turns out money doesn’t grow on narratives.

Tech firms are spending hundreds of billions of dollars on advanced chips and data centers. That’s fine.

The presumption is that AI will soon fundamentally disrupt our way of working and living, so, of course, businesses – and the governments supposed to regulate them – need to prepare. Besides, a lot of the financing is traditionally coming from venture capital and debt.

What isn’t fine is the fact that the money is just crazy. Since the release of ChatGPT in 2022, the value of America’s stock market has risen by $21 trillion, and just ten firms, including Nvidia and Amazon, account for 55% of the rise.

Gartner also estimates that global spending on AI will likely reach a whopping $1.5 trillion by the end of 2025. In January, OpenAI announced a $500 billion AI infrastructure plan known as Stargate.

This doesn’t look natural or healthy, so now, Silicon Valley is urgently debating whether AI firms are actually massively overvalued.

As Bloomberg recently put it, “never before has so much money been spent so rapidly on a technology that, for all its potential, remains somewhat unproven as a profit-making business model.”

What’s more, eyebrows are being raised over some unconventional arrangements, specifically between OpenAI and Nvidia. Is the global chipmaker simply propping up the famous AI firm so that it could keep spending on Nvidia’s chips?

Sam Altman’s baby – which has never turned a profit – is in fact involved in many complex financial deals experts are now quick to call “circular financing.”

They say all this is clouding perceptions of AI demand, and the latter, it seems, isn’t that great. Maybe, just maybe, AI is not actually an industry but a narrative, albeit a truly expensive one?

Murky deals between Nvidia and OpenAI

Warnings about the AI craze are now pouring in. Last week, as per Reuters, the Bank of England warned of an intensifying risk of a “sudden correction” in global markets, adding: “Equity market valuations appear stretched, particularly for technology companies focused on AI.”

A few days later, the International Monetary Fund also said that the AI investment boom might be an economic bubble that could burst, comparable to the dot-com bust in the early 2000s.

Why? Because, simply put, reality is not a fantasy. For instance, Bain & Co said in its September report that AI companies will need $2 trillion in combined annual revenue to fund the computing power needed to meet demand by 2030 – but their revenue is likely to fall $80 billion short of that mark.

The International Monetary Fund said that the AI investment boom might be an economic bubble that could burst, comparable to the dot-com bust in the early 2000s.

Torsten Slok of Apollo, a private investment firm, says that AI stocks are now more richly valued than dotcom stocks in 1999, and UBS adds: “Valuations in the space are indeed flashing red and leave little room for cashflow disappointments.”

Even Sam Altman, OpenAI CEO, who imagines himself as the chief apostle of SAI, is seemingly worried. In August, he said, “Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes.”

It’s nice, this honesty. But isn’t OpenAI the one company behind the craze? The firm has been involved in quite a few interesting, to say the least, financial arrangements with fellow industry players.

Last month, OpenAI made a $100 billion deal with Nvidia, the most valuable publicly traded company in the world. The natural expectation is that the firm will build data centers powered with Nvidia’s chips.

Some analysts say that the chipmaker is clearly trying to prop up its customers – like OpenAI – so that they keep spending on its products. Experts call this “financial engineering” or “circular financing.”

Silicon Valley insiders are also increasingly mentioning Nortel, the Canadian telecom equipment-maker that borrowed prolifically to help finance deals for its customers – and thereby artificially boost demand for its product.

“When a provider invests in an AI company that then commits to spending money on that same provider’s infrastructure, it creates the illusion of organic demand. That makes it harder to distinguish genuine growth and healthy market competition from contractual obligations,” Roman Eloshvili, founder of ComplyControl, provider of AI-driven solutions for banks and financial institutions, told Cybernews.

“Investors see ‘growth’ and assume it reflects customer adoption. When in reality, much of it is artificial inflation, not new money entering the market from real users.”

Is AI actually useful to companies?

Wise words, and the numbers prove it. The Economist says that total revenues of the West’s leading AI firms are now around $50 billion a year.

AI’s potential might be revolutionary, but monetizing it seems difficult for now.

That only seems like a lot, but it’s still a tiny fraction of the $2.9 trillion that companies are forecasted to invest in new data centers between 2025 and 2028.

In other words, AI’s potential might be revolutionary, but monetizing it seems difficult for now.

OpenAI is losing about three times more money than it's earning, and 95% of those using ChatGPT, which generates roughly 70% of the company's recurring revenue, aren’t paying a dime to help stem the losses.

Sure, AI revenues could continue to grow quickly, but only if firms continue to believe the tech is useful to them.

And that’s another glaring issue. Investors were rattled in August when a study by researchers at the Massachusetts Institute of Technology found that 95% of organizations are getting “zero return” from investments in generative AI.

The explanation could be very simple – it’s a case of employers just buying tools like Copilot or ChatGPT for their workers and hoping for the best. Well, they’re definitely getting less than expected, a more recent study says.

According to researchers at Harvard and Stanford, employees are apparently just using AI to create workslop, which is “AI-generated work content that masquerades as good work but lacks the substance to meaningfully advance a given task.”

AI helps workers perform meaningless tasks more efficiently – and more erroneously, that is to say.

Up to 40% of US workers have received workslop from their peers in the past month, the study said, and those who have reported annoyance and confusion, with many perceiving the person who had sent it to them as less reliable, creative, and trustworthy.

This mirrors prior findings that there can be trust penalties to using AI. This also undermines the idea that AI will boost output – it doesn’t because workslop costs millions of dollars a year in lost productivity.

Do we even need to worry about bubbles?

Early AI entrepreneur Jerry Kaplan who founded Go Corporation even mused recently that if the AI hype suddenly dies, the pain will be much more intense than during the dot-com boom. There’s just much more money on the table.

“When the bubble breaks, it’s going to be really bad, and not just for people in AI. It’s going to drag down the rest of the economy,” Kaplan said.

And yet, even though OpenAI is losing money, has bombed GPT-5, and doesn’t actually have $1 trillion it has pledged to spend in the next five years, Altman is adamant that his company – and indeed most of the sector – is in no trouble at all, telling BBC: “There’s something real happening here.”

According to Altman, in order to meet the opportunity, the world actually isn’t spending enough on AI. Even if we’re in a bubble, that’s not necessarily a bad thing, he seems to think.

Plenty agree. Mahdi Yahya, CEO of AI infrastructure start-up Ori, told Cybernews: “Right now, firms are focused on growing up, pushing the limits of intelligence and capability. But once that foundation is built, the focus will shift to growing out – becoming useful to as many industries and companies as possible.”

Sure, like in the dot-com era, many companies, even the high-flying ones, will go bust. But the strongest firms will thrive over the long term – just like Amazon and Google.

“The money fueling this wave isn’t mom and pop 401(k)s. It’s trade, venture, and intercompany capital recycling through infrastructure bets,” Alberto Luengo, CEO of Rkive AI, told Cybernews.

There were two big railway bubbles in Britain in the 1840s and the 1860s but the country still has lots of railways. In America, investors lost a lot of money in electric-light companies, but electricity still came and conquered.

“The tech is real, the delay is just a mismatch between hype and adoption. Most large players are prioritizing long-term foundational layers over short-term applied products.”

History’s on Altman’s side – innovations have indeed more often than not been accompanied by bubbles but progress still happened.

For example, there were two big railway bubbles in Britain in the 1840s and the 1860s but the country still has lots of railways. In America, investors lost a lot of money in electric-light companies, but electricity still came and conquered.

“AI will keep improving and getting cheaper. The winners will be the ones who actually solve real problems with sustainable processes – not just those slapping ‘AI’ on weak products,” Nick Davidov, co-founder and managing partner of DVC, an AI-focused venture fund, told Cybernews.

Unlock more exclusive Cybernews content on YouTube.