The growth and performance of AI-driven investing
What’s the research? A 2026 National Bureau of Economic Research working paper by Shuang Chen, Clemens Sialm and David X. Xu examines the growth and performance of artificial-intelligence-driven investing in US asset management. The authors combine investment advisers’ regulatory disclosures, labour-market data, archived fund-strategy descriptions and Hedge Fund Research data. Their definition is deliberately broader than recent generative AI: AI-driven investing is treated as a specialised subset of quantitative strategies that applies AI technologies to predictive modelling and trading-signal generation, including both autonomous strategies and strategies in which AI signals inform investment decisions.
Which funds are identified? The paper studies 7,896 US hedge funds between 2006 and 2024 and identifies 89 funds—1.1% of the sample—as AI funds under its strict text-based classification. It does not publish a complete named list of those 89 funds. Instead, it reports aggregate fund-level results and strategy characteristics. The authors state that AI funds are concentrated in macro strategies: as of 2024, around 60% fell into the systematic diversified macro category, typically trading liquid instruments such as equity indices, commodities, fixed income and currencies. More than 80% of observed AI-fund assets were allocated to macro strategies. The paper specifically notes that BlackRock’s Absolute Macro Fund was classified as AI-driven from 2021.
Better named examples: Two Sigma InvestmentsDealroom has a profile for this one. Try Dealroom → is a particularly direct example: its own website describes the firm as a quantitative investment and trading firm and says it has used AI in investing for 25 years. Man AHLDealroom has a profile for this one. Try Dealroom → is a long-running systematic manager whose official history lists machine learning as an investment milestone in 2014, alongside automated electronic trading and multi-strategy quantitative programmes. WorldQuantDealroom has a profile for this one. Try Dealroom → describes its quantitative asset-management business as built around data, talent and prediction. These are stronger public examples of AI-/machine-learning-enabled quantitative investing than generic references to established quant firms. The paper also cites Renaissance Technologies and D.E. Shaw as historical practitioners, but does not establish that either is among its 89 text-classified AI funds.
What does it find? AI funds outperformed non-AI hedge funds by approximately 6% annually on a benchmark-adjusted basis in the earlier part of the sample, but the advantage declined over time and became statistically indistinguishable from zero after 2017. The result is not explained simply by early adopters losing an advantage: even AI funds launched in the early years saw most of their relative outperformance disappear after 2017. Comparing AI funds with non-AI sibling funds managed by the same adviser, the authors still find an advantage of approximately 34.9–41.1 basis points per month. AI funds also show lower return comovement than non-AI peers within the same strategy category, contrary to the concern that algorithmic approaches necessarily produce more homogeneous trading. The study therefore presents AI as a source of potential alpha and diversification, but not as a permanent performance guarantee. In 2024, AI-fund assets reached approximately $12bn in the authors’ sample.
Read more: NBER working paper · full paper PDF · Two Sigma · Man AHL · WorldQuant · original X post
