AI bubble risk

J.P. Morgan’s top strategist warns that the artificial intelligence investment boom has expanded so broadly across asset classes that conventional diversification strategies no longer shield portfolios from AI-trade volatility. Investors bullish on AI’s long-term prospects still need to rethink position sizing, leverage, and allocation to avoid outsized drawdowns.

• AI capital expenditure effects now span equities, fixed income, and private markets simultaneously, limiting escape routes.
• Summer momentum unwind in July and August hit AI-linked stocks hardest, exposing portfolio construction risks for AI bulls.
• True diversification remains available only in treasuries, gold, core real estate, and European equities, according to J.P. Morgan analysis.

• July-August Period when momentum unwind hit AI-linked stocks hardest before spreading
• Four asset classes Where genuine diversification still works, per J.P. Morgan testing

Gabriela Santos, chief market strategist for the Americas at J.P. Morgan Asset Management, said the scale of the artificial intelligence buildout has grown so large that finding true portfolio diversification away from the AI trade is now exceptionally difficult. The capital expenditure cycle has spread its effects across nearly every major asset class, from equities and fixed income to private markets, leaving few places for investors to hide.

The concern reflects a structural shift in market dynamics. Over the past 18 months, artificial intelligence has transitioned from a niche technology story to the dominant investment narrative across global markets. Major technology companies and semiconductor firms have committed hundreds of billions of dollars to building out AI infrastructure, data centers, and computational capacity. This capital intensity has rippled through interconnected financial markets in ways that traditional diversification frameworks were not designed to address.

Summer Losses Expose AI Portfolio Concentration Risk

The summer of 2024 delivered a sharp lesson in AI-trade concentration. Stocks linked to artificial intelligence suffered the steepest losses in July, with momentum unwind continuing through August. The episode underscored a critical vulnerability for investors positioned heavily on AI’s narrative momentum rather than fundamentals.

During this period, many portfolios that investors believed were properly diversified experienced correlated declines across seemingly unrelated holdings. The weakness spread from semiconductor manufacturers to cloud infrastructure providers to equipment suppliers and beyond. This synchronized decline revealed that the traditional boundaries between sectors and asset categories had partially dissolved in the face of the AI investment cycle.

Santos stressed that even investors who remain deeply convinced of AI’s ability to drive an extended earnings cycle must adjust their approach to portfolio construction. The risk of oversized drawdowns exists independently of whether AI delivers on its long-term economic promise.

You can be really really bullish AI and still need to think really really carefully about portfolio construction.

Gabriela Santos, Chief Market Strategist for the Americas, J.P. Morgan Asset Management

That recalibration means reassessing position sizing, managing leverage levels, and strengthening diversification across holdings.

AI Buildout Fractures Traditional Sector Groupings

One complication facing portfolio managers is that the shape of the AI buildout continues to shift. Hyperscalers, chipmakers, and software companies, once expected to move in lockstep as a unified trade, now diverge sharply within their own sectors. This breakdown of traditional sector correlations makes historical groupings less reliable for diversification purposes.

The fragmentation stems from uneven capital allocation decisions within the AI ecosystem. Some companies are seeing accelerating orders while others face delays or shifting priorities. Competition for market share in AI services has also intensified, creating winners and losers even within segments that appear homogeneous from the outside.

The concern echoes warnings from other prominent voices on Wall Street, where observers have noted that market behavior has increasingly come to resemble a single, monolithic AI trade. When momentum shifts, the entire complex can respond in unison, magnifying volatility for investors who assumed they had hedged their exposure.

Treasuries, Gold, And European Equities Remain True Hedges

J.P. Morgan tested how closely various assets and portfolios track the broader AI trade by building a dedicated AI factor basket. The results confirmed Santos’s concerns: most assets now move together. Genuine diversification is mostly limited to treasuries, gold, core real estate, and European equities, a stark narrowing from historical norms.

This scarcity of working hedges echoes broader concerns about a stock-bond diversification collapse. Historically, bonds have reliably cushioned portfolios when recessions arrived. For two decades following the 2008 financial crisis, low yields meant bonds alone could perform that protective role.

However, Santos said that dynamic has fundamentally changed. Capital competition has intensified alongside supply shocks, inflation pressures, and rate volatility. To round out positioning in the current environment, investors now need additional inflation-resistant assets beyond traditional stocks and bonds. The shift reflects not just the AI cycle itself but the underlying macroeconomic backdrop of higher structural rates and persistent inflation expectations.

European equities offer exposure to companies less directly dependent on the AI capex cycle, while treasuries and gold continue to function as traditional safe-haven assets during periods of market stress. Core real estate, particularly inflation-linked properties and land, provides inflation protection without the direct AI trade exposure that dominates U.S. equity markets.

Whether this revised diversification mix holds may ultimately depend on how artificial intelligence-related capital spending evolves through the remainder of the year and into 2025, according to Santos’s assessment. If AI investment spending continues to accelerate and spread into additional sectors, even these supposedly safe havens could face unexpected correlations.