AI Skepticism Grows as Market Loses $1 Trillion Amid Bubble Fears

The mounting skepticism around artificial intelligence is validating long-held concerns from critics such as Gary Marcus, a cognitive scientist and longtime AI researcher. For years, Marcus has warned about the overhyped promises of generative AI and the economic risks of a potential bubble. Recent developments, including the lukewarm reception of OpenAI’s GPT-5 and a broad market reaction, appear to be aligning with his caution. The AI hype is now showing signs of waning as companies report high failure rates for AI pilots and investors recalibrate expectations [1].

The AI backlash took shape in late 2024 and early 2025, fueled in part by the MIT study revealing that 95% of corporate generative AI pilots were failing. This prompted a sell-off in tech stocks, with the S&P 500 losing $1 trillion in value as fears of a new dotcom-style bubble intensified. Gary Marcus attributed this market turbulence partly to the underwhelming performance of GPT-5, which many had expected to deliver a significant leap toward artificial general intelligence (AGI) [2]. The disappointment has acted as a wake-up call for investors and tech enthusiasts alike, echoing the kind of overexcitement Marcus has warned about since 2019 [3].

The narrative of a potential AI bubble is further supported by the growing gap between market valuations and actual earnings. Apollo Global’s chief economist, Torsten Slok, noted that the top 10 companies in the S&P 500 were more overvalued than during the dotcom era, with forward P/E ratios and massive market caps appearing detached from fundamentals [4]. The surge in data center investments—projected to hit $3 trillion globally by 2029—adds to the speculative fervor, with such spending rivaling consumer spending’s contribution to GDP [5].

The shift in tone from optimism to caution has also been reflected in public discourse. Prominent figures like former Google CEO Eric Schmidt, once a vocal proponent of AGI being “right around the corner,” have revised their views. In an August 2025 New York Times op-ed, Schmidt co-authored a piece acknowledging the uncertainty surrounding the timeline for achieving AGI [6]. This marks a significant departure from earlier narratives that positioned AI as an imminent transformative force [7].

Despite the growing skepticism, Wall Street remains cautiously optimistic. Banks like Morgan Stanley and UBS have highlighted the long-term efficiency gains and monetization potential from AI, particularly in customer service and enterprise tools. However, analysts like Savita Subramanian of Bank of America warn that the asset-heavy shift toward data centers could mark the end of the asset-light, high-margin model that has driven tech success, potentially warranting lower stock multiples [8].

Marcus himself does not frame himself as a “Cassandra” but as a realist who foresaw the risks. He argues that the human tendency to anthropomorphize AI systems has fueled unrealistic expectations, and that the current overvaluation is in part due to a lack of understanding of the technology’s limitations. “It’s almost tragic,” he said, describing how people project intelligence onto machines that simply do not work like humans [9].

Historical patterns suggest that technological revolutions—railroads, computers, the internet—follow a similar arc: initial euphoria, followed by a crash and then a period of consolidation and real value creation. As data center spending continues to surge, some analysts see this as an inevitable part of the innovation cycle. The question now is not whether AI will deliver long-term value, but whether the current market is overestimating the speed of that realization.

Sources:

[1] https://fortune.com/2025/08/24/is-ai-a-bubble-market-crash-gary-marcus-openai-gpt5/?itm_source=parsely-api