This paper proposes a novel measure of investment companies’ reliance on generative AI, focusing on its implications for hedge funds. We document a sharp increase in generative AI usage by hedge funds after ChatGPT’s 2022 launch. A difference-in-differences test shows that hedge funds adopting generative AI earn 3-5% higher annualized abnormal returns than non-adopters. We further identify this effect by exploiting ChatGPT outages as exogenous shocks. The outperformance originates from funds’ AI talent and ChatGPT’s strength in analyzing firm-specific information. Non-hedge funds yield no significant returns from AI adoption, suggesting generative AI may widen existing disparities among investors.

Mutual funds create liquidity for investors by issuing demandable equity shares, whose value is sensitive to information about illiquid portfolio assets. We study the implications of this information-sensitive liquidity creation by examining frequent trading suspensions in China. These suspensions render stocks illiquid, causing significant mispricing of mutual funds through inaccurate valuations of their illiquid holdings. We find that investors actively acquire information about suspended stocks held by mutual funds. This information drives flows into underpriced funds and is incorporated into stock prices when trading resumes. Our findings suggest that mutual fund liquidity creation stimulates information acquisition about illiquid, information-sensitive assets.

Households invest heavily in actively managed mutual funds with the help of financial advisors. We introduce non-alpha services in the Berk and Green (2004) model to explain the underperformance of broker-sold mutual funds relative to passive benchmarks. Using a dataset covering 3,000 financial advisors and their clients' portfolios over a 13-year period, we find that the most profitable (largest) broker-sold funds pay a significant number of financial advisors to serve a wide base of smaller clients. Regression analysis reveals that fund revenues are primarily driven by the total volume of non-alpha services—measured by the number of advisors serving clients and the number of clients—and by the usage of non-alpha services per client, proxied by the number of financial plans created for each client. In contrast, investment performance (a proxy for alpha services) has a negligible impact. The average client-advisor relationship lasts more than seven years, and advisor assets mostly stem from long-term clients who steadily increase their investments and usage of services. This evidence suggests that clients value non-alpha services more over time.