News on Thursday that ChatGPT maker OpenAI was on track to miss its $20 billion annual revenue target sent stock markets tumbling. The tech-heavy Nasdaq index was hit hardest, falling 1.25% at market close.
Sure, the stock market thankfully rebounded to a near-record high as soon as markets opened Friday, but the whole episode underscores how fragile the AI spending boom is — and how little information we actually have from private AI labs, which have no obligation to keep the public informed.
Perhaps shaken by the momentary crisis, the producers of Bloomberg TelevisionThe daily show Open Interest featured Joachim Klement, managing director and research analyst at investment bank Panmure Liberum.
As a financial analyst, Klement’s view on the AI economics is particularly bleak: In May, he wrote an article in Reuters arguing that institutional investors should not be wondering “if the AI boom will end, but what will happen to the markets when it does.” Now his live interview on Bloomberg‘s flagship news channel highlights how the AI bubble is becoming impossible for mainstream commentators to ignore, even as some publications cite tech billionaires who insist the AI bubble is a myth.
“My fundamental belief is that the AI bubble will burst either in 2027 or 2028,” Klement said in his Bloomberg appearance. This belief is based on the belief that the US AI bubble is focused on the wrong kind of technology: namely resource-intensive frontier models.
“I think the whole AI boom is investing in the bad future of AI,” Klement said. “Right now, we’re investing in large-scale language models running in data centers, while it’s becoming increasingly clear to me that the real future of AI lies in small-language, open-weight models running on local desktops. And that, to me, indicates that we’re already overinvested in data centers, as more and more of these applications migrate to smaller, cheaper models running locally.”
“I think we have overinvested and we will find that these profits,” Klement continued, “that some people – who will remain anonymous at this point – expect to drive up prices, and that profits rise to prices, will not materialize.”
When Bloomberg Presenter Dani Burger raised the elephant in the room – the billions of dollars of data center infrastructure planned for the next few years – Klement explained that the entire investment ultimately depends on a few tech companies whose big plans may never come to fruition.
“As for underlying demand, if you exclude demand from OpenAI and Anthropic, there is virtually no demand for AI computing that justifies the multibillion-dollar investments that hyperscalers alone are expected to make next year,” Klement responded.
To call Klement’s predictions “gloomy” would be an understatement. With the fate of the U.S. economy now completely tied to the data center boom, any slowdown in AI spending will almost certainly be a disaster for everyone, not just the tech industry.
“If you had asked me two months ago, I would have said ‘this is a bubble that can easily last another two years,'” Klement said. Now, “all I need for this bubble to burst is for growth plans to be scaled back and growth to slow, because that will already change the earnings outlook for virtually the entire supply chain.”
Learn more about AI: AI creates ‘doom loops’ throughout the economy
Gn bussni
