BofA Warns AI Bubble Nearing Peak: Bull Market Hinges on Elections and Bond Selloff
0xBroomberg
BofA chief strategist Michael Hartnett warns that AI stocks now make up nearly 39% of the S&P 500 — approaching a concentration ceiling unseen since the 1880s railway mania — and the forces most likely to end this bull run are not fundamentals but November's midterm voters and bond vigilantes.
How concentrated has the market become?
AI-related companies now account for roughly 39% of the S&P 500, approaching the all-time concentration ceiling outside the 1880s railway bubble.
This means → the bubble is about concentration, not rotation. During the 1998–2000 dot-com peak, only tech consistently outperformed; in the final six months before the bust, only tech and telecom posted positive returns.
In plain terms = the entire market's strength rests on a handful of mega-caps. If sentiment flips, there is no cushion from other sectors.
How extreme are the fund flows?
$126.4 billion poured into equity funds in a single week — $119.2 billion into U.S. stocks and $19.2 billion into tech alone, both all-time records.
Annualized, 2026 U.S. equity inflows are on track to hit a record $739 billion. U.S. household equity wealth has already risen roughly $6 trillion this year, creating what Hartnett calls a "wealth–price spiral."
Yet BofA's Bull & Bear Indicator has climbed to 9.2, deep in "extreme bullish" territory and flashing a "sell" signal. This means → historically, global equities have fallen 2%–3% on average within two to three months of such a signal, with drawdowns reaching 15%–20% — though the hit rate is only about 60%, so it is a guide, not gospel.
What is the "votes" risk?
Hartnett's first downside risk is political: if Republicans lose the Senate in November, it could trigger a triple sell-off — dollar down, yields down, stocks down.
The end of the U.S.–Iran conflict has temporarily steadied Trump's approval on economic and inflation issues, but if his numbers fail to rebound by September, bullish sentiment may start to crack.
This reflects Hartnett's broader thesis: historically, booms and bubbles are ended not by deteriorating fundamentals but by voters, bond-market vigilantes, and volatility shocks such as the yen or won crises.
Why are bond vigilantes dangerous?
Bond vigilantes — investors who sell government bonds to "punish" loose fiscal discipline — are already in motion. Since Kevin Warsh was nominated for a Fed post, the 2-year/10-year spread has collapsed from roughly 75 basis points to 25 basis points, a sharp bear-flattening.
This means → the bond market is pricing in "higher inflation equals rate hikes." Hartnett notes that when unemployment falls below the year-on-year CPI — a rare combination — yield-curve inversions and risk events have historically followed.
His only "antidote" is oil: only a significant drop in crude that pushes CPI back below 3% can prevent the narrative from sliding from "inflationary boom" to "stagflationary bust."
What is happening in Asia?
Riding the semiconductor wave, South Korean equities have surged roughly 130.6% year-to-date and Taiwan roughly 63.2%, leading global markets.
Yet Korean assets are exhibiting what Hartnett calls the "ultimate K-shape" — KOSPI keeps climbing while the won has weakened about 5.1% against the dollar. In plain terms = equities and the currency are moving in opposite directions; one of them has to be wrong.
Chinese equities have seen net outflows for a 12th straight week ($9.1 billion in the latest week); European equities for a 10th. Capital is concentrating into U.S. and tech assets at an unprecedented pace.
How should investors read this report?
Hartnett is not calling an imminent crash. He is describing a late-cycle bubble in which returns are hyper-concentrated and sensitivity to political and rate shocks is extreme.
In his framework, whether the bull market survives depends on three things: whether oil prices cool, whether Trump's approval recovers, and the outcome of November's elections.
Important context: the *Flow Show* report has a well-known contrarian, cautious lean. Its sell signals have a historical hit rate of roughly 60% — useful as one input among many, not as a sole trigger.
Content is for reference only, not financial advice.