Breaking

JPMorgan Says Its AI Investing Agents Beat a 60/40 Portfolio in 20-Year Backtests

July 11, 2026 at 14:24 EDT

AI Agents
Finance

JPMorgan Chase has built eight AI-powered investing agents that outperformed a traditional 60/40 stock-and-bond portfolio across roughly two decades of historical market data, according to a bank research note that marks one of Wall Street's most concrete tests yet of agentic AI in capital allocation.

July 9, 2026 · JPMorgan Cross-Asset Strategy

Eight AI Agents Beat the Classic 60/40 Portfolio — On Paper

JPMorgan built eight agentic AI investors that dynamically shift between stocks and bonds. Across ~20 years of backtests, all eight delivered stronger risk-adjusted returns than the benchmark — but the bank calls it research, not a live product.

8 / 8

agents beat the 60/40 benchmark in backtests

+0.7pp

annualized excess return from the best agent, with lower volatility

~20 yrs

of historical market data used in simulation

The static benchmark vs. the dynamic agent

The 60/40 splits capital in fixed proportions. Each agent classifies the market into one of four regimes and reallocates accordingly.

60%

40%

60/40

fixed split

Stocks

Bonds

shift ⇄

AI agent

regime-driven

Four macro regimes drive the allocation

Goldilocks

Reflation

Stagflation

Risk-off

The agents — built on OpenAI and Anthropic foundation models — also outperformed JPMorgan's existing rules-based regime model, adding value beyond the bank's traditional systematic framework.

The promise

A visible test case that agentic AI can replicate and refine the tactical, regime-based shifts human strategists already attempt — with better risk-adjusted results.

The caution

Simulations, not real trading. Backtests risk overfitting and don't guarantee live performance. The bank remains wary of handing allocation decisions to agents.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.