AI Skepticism Grows as S&P 500 Loses $1 Trillion Amid Bubble Fears

The recent turbulence in the AI market has brought long-standing warnings from critics like Gary Marcus into sharper focus, as signs of a potential bubble gain attention amid growing skepticism about the industry's trajectory. Marcus, a longtime AI researcher and cognitive scientist, has been sounding alarms about the limitations of large language models (LLMs) since 2019 and the potential for a financial bubble since 2023. His concerns are now being echoed in the broader market, where the hype around AI is giving way to caution and realism.

The turning point seems to have come with the much-anticipated release of GPT-5 by OpenAI, which many had expected to represent a significant leap toward artificial general intelligence (AGI). However, the model was described by Marcus as “underwhelming” and “disappointing,” with some observers noting it failed to deliver on the AGI promise. This reaction contributed to a broader market jitters, with the S&P 500 shedding roughly $1 trillion in value amid concerns that AI-driven valuations were inflating beyond economic fundamentals.

Torsten Slok, chief economist at Apollo Global Management, has also been vocal about the risks. While stopping short of declaring an outright bubble, he has highlighted that the top 10 companies in the S&P 500 are currently more overvalued than they were during the dotcom era. He pointed to elevated forward price-to-earnings ratios and massive market capitalizations for firms like Nvidia and Microsoft as signs of detachment from earnings. Additionally, he noted that data center investment—seen as essential to support AI growth—now contributes to GDP in a way comparable to consumer spending, a key economic indicator.

The shift in tone was further amplified by a New York Times op-ed co-authored by former Google CEO Eric Schmidt, who questioned the timeline for achieving AGI. This marked a significant departure from earlier assumptions, particularly from those in Washington and the tech policy community who had positioned AGI as an imminent reality. Political scientist Henry Farrell noted that Schmidt’s statement signaled the unraveling of a “New Washington Consensus” that had previously assumed AI would soon deliver transformative intelligence.

While many market participants are now questioning the pace of AI development, Wall Street banks remain cautious but not alarmist. Morgan Stanley and UBS acknowledge the potential for short-term challenges but emphasize the long-term economic benefits of AI adoption. Bank of America, for instance, sees AI as a driver of productivity and innovation for S&P 500 companies, likening the transformation to past technological revolutions.

John Thornhill of the Financial Times offered a broader historical perspective, referencing Carlota Perez’s theory that technological revolutions are often preceded by speculative bubbles and followed by periods of creative destruction. AI, according to Thornhill, fits this pattern and may eventually lead to a “golden age” of productivity, even if the path is rocky.

Gary Marcus attributes much of the market overreaction to human psychology, particularly the tendency to anthropomorphize machines. He argues that people mistakenly project human-like intelligence onto AI systems, leading to unrealistic expectations and overvaluation. This “gullibility gap,” as he terms it, has fueled the current boom and may be reaching a breaking point as models like GPT-5 fail to meet inflated expectations.

Despite the growing skepticism, some, like Bank of America’s Savita Subramanian, acknowledge the transformative potential of AI while cautioning about the risks of overinvestment. She notes that the shift from asset-light tech firms to asset-heavy infrastructure is a significant change that may affect future profitability and valuation multiples.

In the end, whether AI is in the midst of a bubble or simply undergoing a correction remains an open question. What is clear, however, is that the once-unquestioned optimism is giving way to more measured expectations. As Marcus put it, “It’s almost tragic”—a reflection not just of the market’s overreach, but of the human tendency to mistake complexity for intelligence.

Sources:

[1] Title: Is AI a Bubble? Market Crash, Gary Marcus, OpenAI GPT-5 (https://fortune.com/2025/08/24/is-ai-a-bubble-market-crash-gary-marcus-openai-gpt5/)