BofA Director: AI Bubble Is Fit to Burst

Bank of America's Hartnett issues a post-bubble crash roadmap as ai stocks drag indexes into a third consecutive losing session, with breadth and valuation signals mirroring March 2000.

Lead

Michael Hartnett, managing director and chief investment strategist at Bank of America Global Research, declared the artificial intelligence investment cycle a bubble fit to burst on Wednesday and delivered a structured crash roadmap to institutional clients, drawing direct comparisons to the dot-com peak of March 2000. His warning landed as the Nasdaq 100 slipped 0.2% Wednesday after falling 1.03% in the prior session, extending a three-day stretch in which mega-cap AI names led the broader market lower while rising oil prices and Treasury yields applied additional pressure across risk assets.

What Does BofA's Crash Roadmap Say?

Hartnett's post-bubble framework is grounded in nearly a century of precedent. The prescription: go long bonds and build long exposure in defensive and value sectors that significantly underperformed during the bubble's final advance. Consumer staples, mining, materials, and healthcare are the identified rotation targets, reflecting the patterns that followed the 1929 crash, the Nifty Fifty unwind of the 1970s, and the dot-com collapse of 2000 to 2002.

The framework rests on what Hartnett describes as "long humiliation, short hubris" - a recurring post-bubble dynamic in which the most-neglected segments of the prior cycle become leaders of the next recovery while yesterday's consensus positions deflate.

The roadmap represents a sharpening of a position he first flagged in June 2026, when BofA's Bull & Bear Indicator stood at 8.5. By late August, the indicator had climbed to 9.7, its highest reading since 2021. A separate trigger - BofA's FMS Cash Rule - also fired as cash allocations in fund manager surveys fell to 3.6%, another contrarian sell signal under the bank's methodology.

Why Do the Bubble Metrics Align So Closely With March 2000?

The quantitative case has tightened since summer. Only 21 S&P 500 constituents are recording new all-time highs - one more than the 20 present at the dot-com peak. The remaining 331 stocks in the index trade at least 20% below their own record levels, a degree of underlying deterioration that the headline index number entirely conceals.

The S&P 500's price-to-book ratio has reached 5.3, surpassing the 5.1 recorded at the March 2000 top. Of the 20 stocks that touched record highs in late May, only seven had no direct link to AI infrastructure or AI services. Hartnett has also warned that AI-linked names could push domestic market concentration to 47 to 48% of total U.S. market capitalization - a level that would exceed any single-sector concentration recorded during the dot-com era.

The concentration is not a byproduct of broad economic strength. It is the product of a shrinking cluster of AI capex beneficiaries carrying an increasingly heavy share of index returns while the majority of the market quietly retreats.

How Does the AI Capex Complex Amplify Downside Risk?

The structural risk extends beyond valuation. Four hyperscalers - Amazon (AMZN), Microsoft (MSFT), Alphabet (GOOG), and Meta Platforms (META) - have collectively committed hundreds of billions of dollars in data center construction, chip procurement, and compute infrastructure. Nvidia (NVDA), the primary chip supplier to that buildout, has absorbed a disproportionate share of S&P 500 earnings growth and market capitalization gains over the past two years.

Because these names are simultaneously the most liquid, most elevated, and most crowded positions in institutional portfolios, they become the first and fastest assets to sell when macro conditions deteriorate - amplifying drawdowns in ways a more distributed rally would cushion. A chip sector selloff in July 2026 stripped more than $1 trillion in combined market capitalization within a single week, demonstrating the velocity that institutional deleveraging can achieve in concentrated markets.

The additional pressure point: hyperscaler free cash flow has turned negative under the weight of buildout commitments, and return-on-investment timelines for enterprise AI deployment have extended. Financing costs, now moving in the wrong direction as bond yields rise, add further strain to the capex math.

How to Invest in AI Through a Post-Bubble Rotation

Hartnett's framework does not call for abandoning AI as a theme. It calls for distinguishing between the infrastructure build cycle - which has peaked by multiple historical valuation measures - and the monetization cycle, which remains comparatively early. Post-bubble technology recoveries have historically rewarded companies with durable revenue, proven enterprise traction, and valuations supported by cash flow rather than projected capex returns.

The recommended rotation toward bonds, materials, healthcare, and consumer staples compresses risk premium back toward sectors that underperformed during the AI surge. Broad index vehicles including the SPDR S&P 500 ETF Trust (SPY) and the Invesco QQQ Trust (QQQ) reflect the bifurcation between index-level performance and the deteriorating breadth underneath it.

Outlook

Hartnett's roadmap arrives at a moment when the macro backdrop amplifies its urgency. September is historically the worst calendar month for U.S. equities, and the combination of elevated Treasury yields, climbing oil prices, and geopolitical tensions near the Strait of Hormuz provides exactly the category of external catalyst Hartnett flagged in June as capable of puncturing investor complacency. Whether the AI cycle follows the dot-com template depends on whether hyperscaler earnings revisions can justify current multiples before sentiment shifts. BofA's indicator suite suggests the margin for disappointment is running thin.