The Artificial Intelligence (AI) party is in full swing, with tens of billions invested in infrastructure, startups and talent.

Among the most notable announcements this year is the news that Open AI, Softbank and Oracle have committed to investing $500 billion in AI supercomputers.

In China, meanwhile, the giants Alibaba and Tencent increased their injections of resources with the ambition of taking the country to leadership in the field of AI by 2030.

But signs of stagnation are increasingly difficult to miss. Business use of AI is declining. Many economists think concerns about usage, just three years after it became popular, belie the prevailing narrative that AI will revolutionize the way businesses operate by streamlining repetitive tasks and improving predictions.

“The big commitment to infrastructure presupposes a vertiginous increase in its use. However, multiple surveys show that this has decreased since the summer,” Carl-Benedikt Frey, professor of AI at the University of Oxford, tells DW. “Unless new durable utilities emerge soon, the bubble could burst.”

The U.S. Census Bureau, which surveys 1.2 million U.S. businesses every 15 days, found that the use of AI tools at firms with more than 250 employees fell from nearly 14 percent in June to less than 12 percent in August.

The biggest challenge of AIs is their tendency to hallucinate, that is, to generate plausible but false information. Other weaknesses are its reliability and the poor performance of autonomous agents, which complete their tasks successfully only a third of the time.

“Unlike a practitioner who learns on the job, pre-trained AI systems do not improve through experience. We need continuous learning and models that adapt to changing circumstances,” says Frey.

Unsustainable consumption of capital

As the gap between exorbitant expectations and business reality widens, investor enthusiasm for AI is starting to fade. In the third quarter of the year, venture capital deals with private AI companies fell 22 percent quarter-on-quarter.

“What disturbs me is the magnitude of the investment compared to the revenue generated by AI,” economist Stuart Mills, a senior fellow at the London School of Economics, told DW.

Market leader OpenAI generated $3.7 billion last year, against total operating expenses of up to $9 billion. The company says it is on track to earn about $13 billion in revenue this year, but is expected to spend $129 billion by 2029.

Few have quantified the AI ​​bubble as forcefully as Julien Garran, partner at the British firm MacroStrategy Partnership. He estimates that the sheer volume of capital flowing into AI—despite little evidence of sustainable profitability—dwarfs previous speculative frenzies. “It’s 17 times bigger than the dot-com bubble burst,” he told DW.

Investors increasingly cautious

Recent results from Big Tech have generated cautious optimism, but also new doubts about the potential of AI. Data analytics platform Palatir’s third-quarter revenue rose 63 percent, but its share price fell 7 percent following the news. AMD and Meta also saw their strong AI-related results overshadowed by market concerns about system sustainability.

That disconnect between rising values ​​and shaky fundamentals is exactly what worries Mills, who sees a growing gap between what AI promises and what it actually delivers to the market.

When will the bubble burst?

“With the exception of Nvidia, which is selling in droves, most generative AI companies are wildly overvalued,” Gary Marcus, professor of psychology and neuroscience at New York University, told DW. “I estimate that everything will collapse, possibly soon. The fundamentals, both technical and economic, do not make sense.”

On a less gloomy note, Sarah Hoffman, director of AI Thought Leadership at AlphaSense, predicts a “market correction” rather than a “cataclysmic bubble burst.”

After a long period of hype, corporate investment in AI will become more selective, explains Hoffman, with a focus moving from “big promises to clear evidence of the effects” of the offering, in order to ensure that “projects generate measurable returns.”

(dzc/ms)