Jeremy Grantham on How to Tell If a Bubble Is About to Burst

What It Covers

Jeremy Grantham, GMO cofounder and bubble historian, examines whether current AI-driven markets constitute a historic bubble, comparing SpaceX's $2.7 trillion valuation at 100x sales to past manias, and explains the specific market signals that have accurately predicted four major bubble collapses since 1925.

Key Questions Answered

• •Bubble identification signal: Watch for high-flying speculative stocks declining in absolute terms while blue-chip indices continue rising. This pattern appeared in 1929, 1972, 2000, and 2021 — each time preceding a major crash. In 2021, Cathie Wood's ARK portfolio fell 35-40% while the S&P powered higher for months before the broader market broke.
• •Mag Seven structural shift: The seven largest tech companies previously held near-monopolies in distinct verticals — Google in search, Amazon in retail, Meta in social. Now all seven are competing directly in AI, each pledging $127–200 billion in annual CapEx. Grantham frames this as a cage fight replacing seven easy monopolies with one brutal war.
• •Railroad bubble as AI analog: The most historically accurate parallel to AI is the 1800s railroad bubble, not the dot-com era. Railroads genuinely transformed civilization, attracted universal investment enthusiasm, collapsed catastrophically, yet the infrastructure survived and delivered on its promise. AI checks every historical bubble criterion: transformative technology, easy money, and mass participation.
• •Bubble timing and client management: GMO lost half its client book in 2.25 years during the dot-com bubble by being early. The practical lesson: continuously present factual data to clients without hype, maintain consistent communication through bull and bear cycles, and distinguish clearly between "market is overpriced" and an explicit "exit now" recommendation — Grantham has issued the latter only twice in 50 years.
• •Value investing in monopoly stocks: GMO's model identified Microsoft as a buy in the 1990s by projecting return on equity and measuring how market dominance slows mean reversion. Stocks with proven price-setting power and low earnings volatility justify higher multiples. The model held Microsoft in its cheapest value decile continuously until July 1999, demonstrating that quality monopolies can clear quantitative value screens.