The artificial intelligence boom, once the darling of Wall Street and Silicon Valley, is confronting an unprecedented wave of skepticism and self-reflection. Since early August, a confluence of disappointing product launches, damning research, and shifting corporate rhetoric has shaken faith in the technology’s near-term capacity to deliver transformative change—or sustained profits.

Once dismissed as a lone Cassandra, cognitive scientist and AI critic Gary Marcus now appears prescient. For years, Marcus sounded alarms about the limits of deep learning and the market’s exuberance. His central warning: Large language models (LLMs) are hitting economic and technical walls faster than supporters will admit, and mass investment will not indefinitely mask modest commercial returns. “I see myself as a realist who foresaw the problems and was correct about them,” Marcus told Fortune.

A String of Setbacks Jolts the Market

The trigger for the current reckoning was the much-anticipated release of OpenAI’s GPT-5, which failed to deliver the revolutionary leap toward artificial general intelligence (AGI) that had been strongly hinted at in the buildup. CEO Sam Altman himself conceded the company had “totally screwed up” the rollout, and went so far as to publicly utter the dreaded word “bubble”—a rare admission for a tech visionary.

As disappointment set in across the industry, a sweeping MIT survey delivered another punch: a staggering 95% of generative AI pilots at large enterprises are failing to deliver tangible business results. The combination of product letdown and empirical evidence upended the AI growth narrative and spooked investors, resulting in a sharp sell-off that wiped nearly $1 trillion in value from the S&P 500’s tech-heavy portfolio in just days.

This loss, while partly reversed after Federal Reserve Chair Jerome Powell struck a dovish tone about interest rates, revealed the market’s nervousness that years of AI investment may have led to overextended valuations reminiscent of the dot-com bubble.

Warning Signs Multiply: Wall Street Voices Concern

Concerns about the sustainability of the AI surge have been building for months. In July, Torsten Slok, chief economist at Apollo Global Management, highlighted that the forward price-to-earnings (P/E) ratios and market caps of tech giants like Nvidia, Microsoft, Apple, and Meta had become “detached from their earnings”—a classic sign of speculative overheating. Slok noted that top S&P 500 tech companies were even more overvalued than during the peak of the late-1990s dot-com era.

Additionally, ballooning capital expenditure on data centers—necessary to train and run AI models—has come under scrutiny. According to analysts, data center investments have contributed as much to GDP growth in 2025 as consumer spending. Industry giants are spending an estimated $750 billion on infrastructure buildouts over 2024 and 2025, with projections pushing global data center spend to $3 trillion by 2029.

Former Google CEO Eric Schmidt, once a vocal proponent of imminent AGI, recently co-authored a New York Times op-ed noting that the timeline for human-level AI is “uncertain.” This reversal speaks volumes about the mood shift at the intersection of tech and public policy. As political scientist Henry Farrell observed, the “New Washington Consensus” built on AGI’s inevitability is now visibly crumbling.

Backlash and Market Psychology

This summer, a mounting backlash is being felt in both popular and business media. Analysts from think tanks to financial press to viral social platforms are warning that the “AI gold rush” may have overshot reality. The term “AI slop” is trending to describe flawed or overpromised products, while “clunker” has entered tech vernacular for failed automation efforts, particularly in customer service.

Darrell West at Brookings and commentators like Axios and Fast Company all predict that a wave of public and industry skepticism could drive regulatory, investment, and strategic realignments. The “vibe shift” is real: instead of seeing AI as miraculous or magical, many are coming to terms with its limitations—and costs.

History Repeats: Creative Destruction Ahead?

Financial historians urge perspective. John Thornhill of the Financial Times and economic thinkers cite the cycle of technological enthusiasm, overinvestment, speculative bubbles, and eventual creative destruction that has characterized railroads, electrification, the internet, and now AI. Carlota Perez’s landmark book Technological Revolutions and Financial Capital and Edward Chancellor’s Devil Take the Hindmost document how these cycles, while painful in downturns, ultimately yield enduring economic transformation. The current phase, then, may merely be the prelude to a future “golden age” of AI—after a substantial market shakeout.

Supporting this thesis, Owen Lamont at Acadian Asset Management marks the present moment as one when market participants universally know prices are unreasonable—but expect them to rise anyway, a classic bubble dynamic.

Wall Street: Cautious Optimism, Not Panic

Despite turbulence, major Wall Street banks are not universally decrying the AI sector as a bubble. Morgan Stanley projects enormous efficiency gains, estimating up to $920 billion in annual cost savings for S&P 500 companies through AI automation and agentic robotics. UBS foresees “capex indigestion”—a glut of spending on hardware and infrastructure—but maintains that AI adoption is happening more broadly and rapidly than even the optimists predicted, with revenue streams from OpenAI’s ChatGPT and Alphabet’s Gemini growing steadily, as are AI-powered CRM and business systems.

Bank of America’s research makes a nuanced case, emphasizing AI’s role in boosting corporate productivity and enabling businesses to weather inflationary cycles. Savita Subramanian, head of U.S. Equity Strategy, is not ready to call it a bubble for top-tier tech, but warns of inflated valuations among smaller companies and private lenders. The shift toward asset-heavy strategies—i.e., massive data center investments—could threaten the traditional high-margin, asset-light tech business model, possibly justifying a market correction in valuations.

Fundamental Disconnects: Is There Real Value?

Gary Marcus returns to fundamental math when pressed: with nearly 500 AI unicorns now valued collectively at $2.7 trillion, the current monetization cannot justify those numbers. OpenAI reportedly generated $1 billion in revenue (and remains unprofitable), suggesting the global AI market may be generating no more than $25 billion in annual revenue. For an industry with trillions invested, this disconnect is glaring.

Marcus attributes much of the bubble to human psychological biases, especially the tendency to anthropomorphize AI, overestimate its capabilities, and overlook structural limitations. “People look at these machines and make the mistake of attributing human-like intelligence, when the real progress is far more limited,” Marcus says.

Subramanian echoes this, emphasizing that while she personally finds AI tools useful, “the truth is it hasn’t really changed the world that much yet, but I don’t think it’s something to be dismissed.”

Conclusion: A Reckoning with Reality—And the Future

The AI sector faces a pivotal moment: visionary expectations are being forced to reckon with business realities. While short-term pain—corrections, layoffs, and a culling of overhyped firms—seems imminent, the long-term promise of AI remains. As history has shown, technology repeatedly goes through boom, bubble, and bust cycles before fundamentally reshaping society and the economy.

For investors, technologists, and business leaders, the challenge is navigating this turbulence. Awareness of hype, sound fundamentals, and caution are once again in fashion. As Gary Marcus, Wall Street strategists, and economic historians remind us, the future of AI will depend not on magical thinking, but on persistent innovation, real-world results, and sustainable value creation.