He Invested Through Five Bubbles | Andy Constan on What They Taught Him About AI

Why It Matters

This framing matters because investors and allocators face heightened AI-driven enthusiasm and valuation risk; understanding bubble regimes helps set portfolio strategy, risk limits and avoids the futile search for timing the top. It reframes decisions toward process and diversification rather than conviction about market timing.

Summary

Veteran investor Andy Constan says bubbles are identifiable only as regimes with root causes, escalation events and peaking behavior, but their timing and exact tops are essentially unknowable in real time. Drawing on five bubble episodes from his career, he warns that markets can appear ‘‘overbought’’ while still reflecting consensus prices, and that distinguishing genuine technological breakthroughs (like AI) from bubble dynamics requires framework-based analysis rather than market-timing calls. Constan counsels humility: most investors are better off with passive, diversified positions while recognizing when market regimes shift toward bubble-like behavior. The goal, he says, is to learn how to think about bubbles—not to predict their precise end.

Original Description

First Principles with Andy Constan launches with a deep dive into market bubbles, AI, semiconductor stocks, and the financial conditions that can turn powerful technological change into a dangerous investment regime. Andy explains how bubbles form, why they are almost impossible to time, how today’s AI boom compares to past episodes like 1987, the dot-com bubble, housing, and the bond bubble, and what investors should watch as expectations, financing, and FOMO build.

Andy Constan on X

https://x.com/dampedspring

Damped Spring Advisors

https://dampedspring.com/

Topics covered:

* Why bubbles are easy to identify in hindsight but nearly impossible to define in real time

* The difference between an expensive market and a true bubble regime

* How new technologies, easy money, regulation, and exogenous shocks can create bubble conditions

* Why AI may rhyme with the internet boom without being an exact repeat

* The role of ChatGPT, Microsoft’s OpenAI investment, and semiconductor earnings expectations

* What the 1987 crash, Japan, housing, bonds, and dot-com bubble can teach investors today

* Why human nature, FOMO, and “keeping up with the Joneses” make bubbles so powerful

* How the late-1990s Fed response to Long-Term Capital Management helped fuel the final phase of the tech bubble

* Why tech’s current size in the economy and market may limit how far the AI boom can grow

* How AI capex, hyperscaler spending, buybacks, debt issuance, and IPO supply could determine what happens next

Timestamps:

00:00 Intro and the challenge of identifying bubbles

04:32 Expensive markets vs true bubble regimes

09:57 The five bubble episodes Andy compares to today

14:35 Root conditions, escalation events, and the peaking phase

19:20 Why the 1987 crash may also have been a bubble

24:25 The late-1990s setup and the Netscape Navigator moment

28:00 Crisis analogs, easy financial conditions, and today’s AI parallels

32:20 Long-Term Capital Management and rocket fuel for the tech bubble

36:11 Why tech’s market share matters more today than in the 1990s

43:18 Policy mistakes, subsidies, and how governments feed bubbles

47:42 Semiconductor earnings expectations and valuation risk

53:45 The AI capex chain and where the money has to come from

58:42 IPOs, corporate debt, and the financing risk behind the AI boom

01:02:27 What investors should do differently in a bubble regime