JPMorgan Chase has built experimental AI-driven investing agents that switch between stocks and bonds based on market conditions and, in two decades of backtests, outperformed a classic 60/40 portfolio by 0.7 percentage point annually with lower volatility. The models also beat the bank’s own rules-based regime framework, according to strategists led by Thomas Salopek. The bank stressed the findings are hypothetical and cautioned against assuming the systems can consistently beat markets in real time. Still, the work underscores Wall Street’s push to move AI from back-office aid to capital-allocation decisions—while raising concerns about crowded trades and market stress if firms converge on similar models.
Related articles:
— FinGPT: Open-Source Financial Large Language Models
— BlackRock Aladdin: Technology platform for investment professionals
