Are We Heading for an AI Dot-Com Bubble? A 2025 Update

In 2024, market watchers noted that the performance gap between traditional market indexes and their equal-weighted counterparts was the widest since the peak of the dot-com bubble in 2000¹. The S&P 500 Equal Weight Index underperformed its cap-weighted benchmark by 6.9 percentage points in Q2 2024, the third widest spread since the equal weight index’s inception in 1989². This raised serious concerns of an AI-driven bubble.

A year later, prominent economists, academics, and policymakers argue that AI valuations may be overextended. Others emphasize AI’s real-world applications and productivity gains. Below, we summarize both perspectives with proper sourcing.

AI: A Tool for Industry Innovation

In 2024, we argued that AI’s practical uses set it apart from the speculative dot-com era. That remains true today:

Better Tools: AI is improving productivity across multiple sectors. Research from Stanford and MIT found that customer support agents using generative AI tools achieved 14% productivity gains on average, with less experienced workers seeing improvements up to 35%³⁴. Similar studies document productivity gains in software development and other knowledge work.

Industry Transformation: AI adoption is expanding across finance, manufacturing, and retail, where it is already driving measurable improvements in fraud detection, robotics, and supply chain optimization.

Sustainable Growth: AI platforms are being built into business infrastructure worldwide, enabling organizations of all sizes to innovate and remain competitive.

The Bubble Argument

Several prominent voices warn that AI could be overhyped:

Torsten Sløk (Chief Economist, Apollo Global Management) has been particularly vocal about AI bubble risks. In July 2025, Sløk stated: “The difference between the IT bubble in the 1990s and the AI bubble today is that the top 10 companies in the S&P 500 today are more overvalued than they were in the 1990s”⁵⁶. He noted that the S&P 500 has become “extremely concentrated,” with the top 10 stocks contributing 54% of market returns since January 2021⁷.

Daron Acemoglu (MIT) offers a more measured but cautious view. The 2024 Nobel laureate estimates that AI will produce only a “modest increase” in GDP between 1.1 to 1.6 percent over the next 10 years, with roughly 0.05 percent annual productivity gains⁸⁹. Acemoglu argues that “we currently have the wrong direction for AI. We’re using it too much for automation and not enough for providing expertise and information to workers”¹⁰.

Erik Brynjolfsson (Stanford) takes a more optimistic but realistic stance. His research documented significant productivity gains in customer service (14% average improvement)¹¹, but he acknowledges implementation challenges and the need for organizational change.

Sam Altman (OpenAI) has surprisingly acknowledged bubble concerns. In August 2025, Altman said AI valuations have become “insane” and admitted “Are we in a phase where investors as a whole are overexcited about AI?” while noting that some investors will be left “very burnt”¹².

Jerome Powell (Federal Reserve Chair) has expressed concerns about AI’s economic concentration. In September 2025, Powell noted the U.S. is seeing “unusually large amounts of economic activity through the AI buildout,” with spending that “may well be skewed toward higher-earning consumers”¹³. Powell suggested AI is “probably a factor” in concerning unemployment rates, particularly affecting young graduates¹⁴.

Erik Gordon (University of Michigan) has issued the most dire warnings. Gordon describes the AI surge as an “order-of-magnitude overvaluation bubble” and warns that “more investors will suffer than suffered in the dot-com crash, and their suffering will be more painful”¹⁵¹⁶. He points to CoreWeave’s 33% stock decline in two days, wiping out $24 billion in market cap—almost 60 times Pets.com’s peak market value¹⁷.

These voices raise concerns about market concentration, inflated expectations, and the risk of large infrastructure investments becoming stranded assets if demand falters.

Counterarguments: Why AI is Different

Supporters of AI’s long-term potential counter these warnings:

Real Demand: Unlike the dot-com era, today’s AI hardware and software are already in use, with companies reporting measurable returns. Stanford research shows customer satisfaction improved alongside productivity gains when AI tools were deployed¹⁸.

Diversification: The AI ecosystem includes large firms, startups, governments, and nonprofits, reducing reliance on any single player.

Healthy Corrections: Market data shows significant rotation and corrections have already occurred. In 2024, while the Magnificent Seven gained 37%, the S&P 500 Equal Weight Index gained only 5.1%, and many individual stocks experienced bear-market-level declines even as indexes remained positive¹⁹²⁰.

Policy Support: Public initiatives in the U.S., Europe, and Asia are providing a foundation for long-term investment, in contrast to the dot-com era’s reliance on private speculation.

The Importance of AI Infrastructure

AI infrastructure — especially specialized chips and data centers — remains central to the debate. High barriers to entry limit the number of providers, but new entrants and diversified investments are increasing competition:

High Barriers: Advanced design and manufacturing require rare expertise, supporting resilience against oversupply.

Specialized Players: Startups are building chips tailored for specific AI workloads.

Strategic Investment: Many organizations are designing custom systems to reduce reliance on external suppliers, spreading risk across the market.

Market Concentration Data

Current market concentration has reached extreme levels. The top 10 companies in the S&P 500 now account for approximately 39% of the index weight—the second record-high set in 2024. This compares to less than 27% at the height of the dot-com era. For the full year 2024, only 19% of stocks within the S&P 500 outperformed the index itself²¹²².

So what do we think? 

Prominent voices including Torsten Sløk, Daron Acemoglu, Erik Brynjolfsson, Sam Altman, Jerome Powell, and Erik Gordon present legitimate concerns about AI valuations, market concentration, and the timeline for meaningful productivity gains.

However, evidence of real productivity improvements, diversification across industries, and policy support suggest the AI boom differs fundamentally from the dot-com era. While Acemoglu’s research suggests modest near-term economic impact, he acknowledges that AI could have much greater effects if development focuses on augmenting rather than replacing workers²³.

The data points to a market that, while potentially overvalued in certain segments, is built on stronger fundamentals than the dot-com bubble. AI is shaping up to be not just another financial cycle, but a lasting transformation in how industries innovate and grow—though perhaps not as quickly or dramatically as some investors expect.

---

Sources

¹ Based on historical performance data comparing cap-weighted and equal-weighted indexes

² NASDAQ First Half 2024 Review and Outlook

³ Brynjolfsson, E., Li, D., & Raymond, L. (2023). “Generative AI at Work.” NBER Working Paper

⁴ Stanford Institute for Economic Policy Research (2023). “Generative AI boost can boost productivity without replacing workers”

⁵ Apollo Academy (July 16, 2025). “AI Bubble Today Is Bigger Than the IT Bubble in the 1990s”

⁶ Fortune (July 17, 2025). “Apollo’s chief economist warns the AI bubble is even worse than the 1999 dot-com bubble”

⁷ Fortune (July 25, 2025). “Apollo economist: Mag 7 may not be the best AI investment as S&P 500 grows more concentrated”

⁸ MIT Economics (2024). “Daron Acemoglu: What do we know about the economics of AI?”

⁹ MIT Sloan (January 21, 2025). “A new look at the economics of AI”

¹⁰ MIT Economics (2024). “Daron Acemoglu: What do we know about the economics of AI?”

¹¹ NBER Working Paper No. 31161. “Generative AI at Work”

¹² CNBC (August 19, 2025). “Analysts downplay AI bubble worries as Altman says some investors will be left ‘very burnt'”

¹³ Fortune (September 17, 2025). “Jerome Powell on signs of an AI bubble and an economy leaning too hard on the rich”

¹⁴ Gizmodo (September 17, 2025). “Fed Chair Powell Says AI Probably a Factor in Concerning Unemployment Rates”

¹⁵ The WealthAdvisor (August 15, 2025). “Could AI Investment Surge Collapse and Inflict Greater Damage on Investors Than Dot-Com Bust?”

¹⁶ Yahoo Finance (August 15, 2025). “Tech guru Erik Gordon says investors will ‘suffer’ far more from the AI boom than the dot-com crash”

¹⁷ Yahoo Finance (August 15, 2025). “Tech guru Erik Gordon says investors will ‘suffer’ far more from the AI boom than the dot-com crash”

¹⁸ Stanford Institute for Economic Policy Research (2023). “Generative AI boost can boost productivity without replacing workers”

¹⁹ NASDAQ (2024). “First Half 2024 Review and Outlook”

²⁰ Charles Schwab (2024). “It Was a Very Good Year”

²¹ NASDAQ (2024). “First Half 2024 Review and Outlook”

²² Charles Schwab (2024). “It Was a Very Good Year”

²³ MIT Sloan (January 21, 2025). “A new look at the economics of AI”