Story
JPMorgan AI Agents Outperform 60/40 Portfolio in Historical Simulations
Summary
JPMorgan Chase reported that AI-powered investment agents outperformed a traditional 60/40 portfolio in historical simulations over the past two decades, though the bank cautioned the results are not from live trading.
Artificial intelligence agents developed by JPMorgan Chase & Co. outperformed a traditional 60/40 portfolio in historical simulations, according to a research note from the bank cited by Bloomberg. The findings represent Wall Street's latest exploration into using AI for complex capital allocation decisions.
Backtest Performance
In backtests spanning the last two decades, researchers built AI agents designed to dynamically shift allocations between stocks and bonds by identifying changing market conditions. The best-performing system reportedly topped a standard 60/40 stock-bond portfolio by 0.7 percentage points annually.
According to the note from strategists led by Thomas Salopek, the AI model achieved this outperformance with lower volatility. It also surpassed the returns of JPMorgan’s own rules-based market regime model. The bank described the work as its first attempt at building an AI system specifically for identifying these market regimes.
A Note of Caution
AdDespite the promising results, JPMorgan's strategists issued a strong warning against over-interpreting the data, emphasizing that the findings are based on historical simulations, not live trading. The note cautioned against "uncritically accepting what amounts to in-sample, overly confident answers of AI."
The researchers stressed that for AI to be effective in asset management, it must be integrated into a well-defined investment process. They wrote that "agentic AI needs to be grounded in a well thought-out asset allocation process, rather than naively assuming the agent can be the source of the domain knowledge."
Wall Street's AI Push
This research highlights a broader industry trend where financial institutions are moving beyond using large language models for research and coding. Banks are now actively testing AI's potential to make sophisticated, real-time investment decisions across different markets, representing a significant potential shift in asset management strategy.