Why Andy Constan Says The AI Bubble is in Earnings, Not Price

Episode Summary

Executive Summary: Andy Constan argues the market is in a bubble, but not because valuations look extreme; instead, he says earnings expectations—especially for AI-linked companies—have become unsustainable relative to the economy’s ability to generate enough GDP to fund them. He believes the AI capex cycle, inflation persistence, and policy inertia create a regime where winners can keep winning, but only by taking share from everyone else.

Main Topics: Why this is a bubble regime, not a normal bull market (Priority: 5/5): Constan distinguishes bubble regimes from ordinary market cycles, saying he has lived through multiple bubbles and now sees a sixth. His focus is regime change, not top-calling, and he argues today’s setup resembles prior bubbles when a new catalyst plus easing conditions sparked parabolic moves. The bubble is in earnings expectations, not price/PE (Priority: 5/5): He says the market’s traditional valuation measures are not yet obviously extreme because earnings estimates have risen sharply. The danger, in his view, is that the E in PE is inflated by expectations that cannot all be met simultaneously across the market. AI capex as a concentrated bet across semis, hyperscalers, and frontier models (Priority: 5/5): Constan describes AI as a massive, concentrated investment cycle involving a few major cohorts that are funding compute-heavy expansion through debt, equity, IPOs, and reduced buybacks. He sees the setup as powerful but not yet proven in terms of ROI. GDP as the ultimate constraint on corporate earnings (Priority: 5/5): A core thesis is that aggregate corporate profits cannot sustainably outgrow the economy’s available pie. He argues AI names may capture more share, but if their earnings rise too quickly, the rest of the market must lose share or GDP must unexpectedly accelerate. Inflation and policy failure keep risk assets supported (Priority: 4/5): Constan stresses that policymakers have not meaningfully defeated inflation and that asset-rich households can keep spending despite weak real-wage growth. This supports higher-for-longer rates and makes a garden-variety recession or easy policy-driven market rollover less likely. How to trade a bubble: discipline, not heroics (Priority: 4/5): He says shorting bubbles is dangerous and that the better approach is disciplined participation, taking profits, and using hedges like collars or options. He references trailing stops and avoiding the mistake of re-entering after a bubble has already popped.

Key Arguments: The market can be in a bubble even if PE ratios are not stretched, because earnings expectations may be the unstable variable. AI-related firms could earn enormous profits, but only by taking a disproportionate share of total GDP, which cannot expand fast enough for everyone to win. The current AI cycle is a concentrated capital bet among semis, hyperscalers, and frontier models, financed through multiple channels and still awaiting proof of ROI. Inflation remains unresolved, and policymakers have done too little to restrain it, which helps keep nominal asset prices elevated. A true bubble pop would likely be far more severe than a routine 2022-style drawdown; it could involve 70% to 80% declines in leading names. The safest way to deal with bubbles is not to predict the top, but to manage exposure with discipline, profit-taking, and hedges.

Data Points: Bubble episodes cited: 6 total - Constan says he has lived through four or five and believes this is a sixth bubble regime Prior bubble period: 1982-1987 - He cites stocks, deregulation, and the end of inflation as bubble conditions Dot-com bubble period: 1995-2000 - He cites Netscape, easing after 1994 tightening, and strong stock market rally Housing/credit bubble period: 2005-2008 - He identifies housing and credit as another major bubble Japan bubble period: 1988-1989 - He references Japan as an additional historical bubble example Government bond bubble period: 2018-2021 - He says bond prices went parabolic after ZERP, the repo crisis response, and COVID Expected GDP growth created: $1.5 trillion - He estimates year-over-year GDP growth available to support corporate earnings Corporate share of GDP growth: 10%-12% - He says corporations typically capture this portion of incremental GDP Implied corporate share of new GDP: $175B-$200B - Derived from his estimate of corporate take from $1.5T of new GDP Public AI stocks expected earnings increase: $250 billion - He says public AI stocks are expected to earn this much more next year S&P 500 total expected earnings: $400 billion - He says the full index is expected to earn this much more AI share of S&P earnings growth: 62.5% - He says consensus estimates imply AI companies will account for 62.5% of all S&P year-over-year earnings change Inflation duration above target: 62 months - He says inflation has been well above target for this long 2022 bear market context: Massive tightening, not bubble pop - He argues 2022 was a policy-driven drawdown, not a true bubble collapse Possible bubble drawdown severity: 70%-80% - He says real bubble pops are typically far larger than ordinary declines Add-on market upside if bubble persists: Up to 60% higher - He says if this is a bubble, it could still rise substantially before topping IPO comparison: 3 deals vs 400 IPOs - He says forthcoming major AI-related IPOs could be as large as the proceeds from roughly 400 IPOs in the late 1990s

Pivotal Quotes: "It's not the P that's the bubble, it's the E." — Andy Constan: He explains that earnings expectations, not valuation multiples, are the bubble-like feature today "There just can't possibly be that way. Somebody has to be the loser." — Andy Constan: He argues AI and the rest of the market cannot all meet current earnings expectations simultaneously "You don't short a bubble." — Andy Constan: He discusses trading discipline and the danger of fighting a bubble with outright shorts

Implications: If Constan is right, AI leaders may keep outperforming, but broader markets face a growing risk that earnings assumptions outrun the economy. Investors should focus on exposure management, hedging, and policy/inflation signals rather than trying to call the exact top.