The Bridge Ep. 11: Expensive Isn't a Bubble

Summary

Cliff Asness draws a sharp line between “expensive” and “bubble,” arguing that a bubble requires valuation extremes that cannot be justified under reasonable assumptions, not just stretched prices. He says markets are currently expensive and concentrated, but still below the extreme dispersion seen at the dot-com peak and the post-COVID valuation spike. The discussion also digs into how AQR builds market-neutral quant portfolios, with roughly 1,000 longs and 1,000 shorts balanced by industry and country, and how modern quant has expanded beyond cheap multiples into profitability, low risk, momentum, and broader valuation. Asness emphasizes that crowding is a real risk because enough capital can compress spreads and lower future returns, while alternative data and machine learning are useful but only if they improve the signal without becoming opaque or overfit.

Notes

• AQR’s core approach is designed to produce returns with low correlation to the market.
• A typical AQR global portfolio is about 1,000 long stocks and 1,000 short stocks.
• The portfolio is mainly balanced by industry and country rather than by single-stock conviction alone.
• Long and short signals are constructed symmetrically; shorts are meant to be bad on the same dimensions longs are good on.
• AQR uses alternative data, NLP, and AI tools, but keeps unique alpha proprietary.
• Early quant research focused on low multiple stocks and momentum.
• Asness rejects small-cap outperformance as a core belief.
• He argues that value is misnamed if it is reduced to cheap multiples alone.
• Paying less is still generally a good thing, especially when paired with momentum.
• Modern quant has evolved into a broader, more diversified DCF-like framework.
• Asness sets a high bar for calling something a bubble.
• A bubble must be beyond-the-pale pricing that cannot be justified by non-ridiculous assumptions.
• He distinguishes between recognizing a bubble and profiting from it; timing is the hard part.
• The market is expensive and concentrated, but he says that is not yet enough to call it a bubble.
• His measure of today’s valuation spread is around the 75th percentile, not prior bubble extremes.
• The dot-com bubble was one of his two major bubble peaks.
• He also identifies the post-COVID peak as a bubble-like extreme.
• At the dot-com peak, expensive stocks traded about 12 to 13 times cheap stocks on his spread measure.
• Historical valuation spreads were usually about 3 to 6 times before the tech bubble.
• Issuance is a soft bubble signal, and he says the market has not yet shown the massive issuance typical of past bubbles.
• Enough capital can ruin any investment strategy by compressing the spread and lowering expected returns.
• Crowding can matter both as long-term capacity compression and as short-term left-tail events.
• The August 2007 quant crisis was a major but brief event for AQR.
• AQR reduced leverage after the quant quake and retired some highly leveraged strategies.
• AQR now relies more on proprietary alternative-data and ML approaches than on its original factor exposures.
• Alternative data refers to newly created or newly assembled databases sold to investors.
• Alt-data edges are often fastest in retail and are most vulnerable to arbitrage.
• He calls alternative data a form of fundamental momentum.
• AQR uses earnings-call NLP as a machine-learning input.
• AQR worries about both overfitting and underfitting in machine learning.
• ML helps AQR build subtler valuation models and measure momentum more effectively.
• AQR has not fully handed investment decisions over to ML because it still wants intuitive validation.