🔥 🏦 JPMorgan tests AI investment agents: Wall Street begins to upend itself

JPMorgan is testing an AI agent that can autonomously adjust allocations between stocks and bonds. The backtest results look promising—over 20 years of history, the best model’s annualized return outperformed the traditional 60/40 portfolio by 0.7 percentage points, with lower volatility. But JPMorgan itself warns that large-scale adoption of AI may lead to convergence of trading strategies, potentially amplifying volatility under stress.
Crypto markets have long grown aesthetically tired of AI narratives, but this time it’s different. JPMorgan is testing investable AI agents with real money. If one of the world’s largest investment banks starts using AI for asset allocation, traditional finance’s “quant” approach is evolving from factor models to an autonomous decision stage. The impact on crypto won’t be in short-term prices, but in structural changes to where capital flows—when traditional assets can also be dynamically optimized by AI, crypto’s appeal as “alternative beta” will be repriced.
Outperforming by 0.7 percentage points in backtests may sound small, but given JPMorgan’s tens of trillions in asset management scale, it’s enough to drive massive capital migration. The real hidden risk is strategy convergence: if most institutions use similar AI models, the market could see one-sided stampedes in extreme conditions, and the liquidity “black hole” from March 2020 may happen again. Crypto markets are already liquidity-fragile, so cross-asset contagion would be even more severe.
AI agents rely on historical data and can’t handle true black swans; strategy convergence may compress volatility, but it can also trigger a concentrated blow-up during stress tests. JPMorgan is candid that this is “based on simulated tests, not actual performance,” suggesting they don’t fully trust AI themselves yet. For crypto traders, rather than chasing the next AI token, consider this: when Wall Street’s AI starts rebalancing on its own, your counterparty may no longer be human—it could be a cluster of algorithms that has no emotions, but will make collective mistakes.
$ai #defi #ai #区块链 #Crypto market