We Are in the ‘Advanced Stages’ of an AI Bubble, says Rockefeller’s Ruchir Sharma
Ruchir Sharma, Chairman of Rockefeller International, has argued that the current AI-driven surge in US tech stocks is in the advanced stages of a bubble.
In a recent Financial Times opinion piece and CNBC interviews he applies his diagnostic framework of four “O’s”—overvaluation, over-ownership, over-investment, and over-leverage—to assess bubbles, concluding that the AI mania strongly exhibits the first three and shows emerging signs of the fourth.
Sharma’s Four “O’s” Framework
Sharma uses this to identify bubbles based on historical patterns:
- Overvaluation – US tech stocks have risen tenfold (inflation-adjusted) over the past 10-15 years, matching major historical bubbles like the 1970s gold rush or 1990s internet boom. Valuations sit in the 95th percentile historically, only exceeded during the 1999-2000 dot-com peak. A century of data shows a >50% crash probability when a hot industry outperforms the broader market by >100% over two years—AI-related stocks approach this threshold. Sharma notes: “if history is any guide, then valuation and prices are flashing a deep-red bubble warning.”
- Over-ownership – US households hold a record 52% of wealth in stocks (higher than the 2000 peak and far above Europe at 30% or Japan at 20%). Daily share trading hit 18 billion (up 60%), with retail options trading rising from one-third to over half of activity. Speculation concentrates in the “Magnificent Seven” tech giants, fueled by young investors and $7.5 trillion in money market funds. Even skeptical institutions feel forced to participate, creating “fully invested bears.” Growing public worry about an AI bubble mirrors pre-dot-com crash concerns.
- Over-investment – Tech spending exceeds 6% of US GDP, topping the 2000 record. The Magnificent Seven’s AI capex doubled to $380 billion in 2025, projected to surpass $660 billion by 2030 (e.g., data centers and power infrastructure). Yet actual AI adoption remains low (<15% of US companies use it meaningfully, with slowing growth). Techno-optimists predict AI will replace 40% of human tasks and cut labor costs, but Sharma warns returns remain unclear and potential job losses (e.g., unemployment to 20%) could spark political backlash.
- Over-leverage – This shows less extremely than in past bubbles, offering some comfort. Big tech firms (e.g., Amazon, Meta, Microsoft) turned net debtors recently after being cash-rich, issuing significant debt for AI. Leveraged ETFs grew sevenfold to $140 billion, amplifying retail bets. Ties to high US government deficits could push interest rates higher if bond markets react.
Sharma’s Overall Conclusion and Caveats
Sharma concludes: “we are pretty much in the advanced stages of a bubble.” He compares it to past manias but calls it uniquely “the most hated bubble in history” due to widespread skepticism and fears over AI’s societal impacts (e.g., job displacement). Bubbles rarely burst on their own; they typically need a catalyst like rising interest rates from tighter monetary policy. The current bubble could persist longer if easy money continues, but historical patterns suggest vulnerability if financial conditions tighten.
