U.S.-China Relations
NextFin News - The summer's "extreme bull sentiment" is over, and Bank of America's Michael Hartnett is telling clients to stay risk-off for the next three to four weeks as a confluence of higher real yields, U.S. midterm election uncertainty, and cracking AI-linked price action pushes investors toward the exits. The strategist's latest note frames the paradox of the moment: investors are simultaneously piling into the most crowded tech trades since 2020 while bracing for a bond-driven deleveraging that could end the AI boom before 2027. The question is no longer whether the AI rally has gone too far — it is what, exactly, pops it first, and whether the Federal Reserve's next move is the pin.
The Setup: A Survey That Contradicts Itself
Bank of America's October Global Fund Manager Survey, conducted October 2-8 among 161 managers overseeing a combined $453 billion, captures the market's central tension in a single contradiction. Sentiment fell to 6.4 from 7.0, its lowest level since June, and net overweight positioning in stocks retreated to 44% from 49%. Yet the proportion of managers running overweight positions in technology rose to 37%, up from 30% a month earlier and at its highest since August 2020. The most crowded trade in the world is "long global semiconductor stocks" at 45%, followed by "short U.S. Treasuries" at 27% and "long Magnificent Seven" at 16%.
That positioning sits on a knife's edge of yield sensitivity. Thirty-seven percent of respondents said they would rotate from stocks to bonds if the U.S. 10-year Treasury yield reaches 6%, another 14% set their threshold at 6.5%, and 19% said they have already begun rotating. The benchmark 10-year yield briefly touched 5.33% on October 14 before settling near 5.3%, having climbed 1.16 percentage points over the past year to roughly 5.2%. Cash allocations held steady at 3.9% of assets — a level that, under BofA's own "cash rule," remains below the 4.0% trigger and therefore still signals a sell for equities, just as it did in September when cash rose to 3.9% from 3.5%.
Weekly flow data confirms the defensive tilt without showing capitulation. Bonds drew $17.3 billion, cash $11.6 billion, crypto $3.0 billion and gold $2.3 billion, while $10.2 billion exited equities, with technology and financials among the weekly laggards. Yet global equities remain on pace for a record $1.2 trillion of inflows in 2026, comfortably above the prior record of $949 billion set in 2021. Investors are trimming risk around the edges of a trade they have not abandoned.
Hartnett's diagnosis is blunt: the broad selloff in global bonds has driven inflation-adjusted real yields to 3%, and that repricing of the price of money is ending the bond bubble that underpinned the AI build-out. The strategist has condensed the thesis into three words in his Flow Show note: "Bond bubble ends." The implication is counterintuitive — and it is the part of the argument most investors are missing.
Why a Fed Rate Hike Could Be Bullish
Conventional wisdom holds that higher rates are bad for stocks. Hartnett's argument inverts it. His team contends that the most bullish catalyst for risk assets before year-end would be a Federal Reserve rate hike in October, because a deliberate tightening move could cap runaway long-end yields and prevent a more damaging disorderly surge in the bond market.
The logic runs through performance dispersion. U.S. stocks are up 13.4% this year while government bonds have fallen 4.4%, underperforming cash by 7.2 percentage points. Stocks have not been knocked down by rates because U.S. hard data, global earnings expectations, and hyperscaler capital-expenditure forecasts have not been meaningfully revised lower. Bonds, by contrast, are already showing cracks at the margin: Korean brokerages, French government debt, and CCC-rated credit have all broken first. Since the start of 2026, global central banks have implemented 54 rate hikes — the first net tightening cycle since September 2023.
South Korea is the canary. Over the past 18 months, Korean 10-year yields have risen approximately 200 basis points. Mirae Asset, the country's largest retail brokerage, is down 65% since May — even as the broader Korean market is up 98.9% this year. The lesson is not that the Korean bubble has burst; it is that when the cost of capital rises suddenly, the most leveraged and most rate-sensitive parts of the market break first. That is the transmission mechanism Hartnett is tracing into U.S. assets: not a broad earnings recession, but a targeted repricing of the duration-heavy, leverage-dependent corners where the AI trade actually lives.
Hartnett has made the deleveraging trigger explicit. If the iShares Global Financials ETF falls below $125 while the MOVE Index — the bond-market's volatility gauge — rises above 125, the combination signals that financial leverage is breaking alongside bond volatility, the classic setup for a broader risk-off event. Those thresholds are not abstract: the MOVE Index jumped roughly a third in two days during the late-September bond selloff, before yields stabilized on strong overseas demand at the 10-year auction.
"Central banks never run out of ammunition, ever, but they can run out of credibility."
That line, from Hartnett's October 1 note, frames the credibility problem now facing the Fed. With 76% of survey respondents expecting the central bank to hold rates steady at the October 28 meeting, the market is pricing in passivity. Hartnett's warning is that passivity could allow long-term yields to spiral, forcing a more violent adjustment later. The bond market, in this reading, is not a passive backdrop — it is the active constraint.
The Midterm Variable: Politics as a Discount Rate
Between the Fed and the bond market sits a third variable: the U.S. midterm elections. Hartnett's team has mapped election outcomes to equity returns. A divided government — Republicans retaining the Senate while Democrats take the House — corresponds to a "Goldilocks" risk-appetite scenario, with stocks potentially rising more than 5%. A Republican sweep would be the largest upside surprise, with stocks potentially rising more than 10% and the AI trade continuing its trajectory toward 2027 bubble conditions. A Democratic sweep, by contrast, could shift the political discount rate applied to stocks abruptly.
This is why the risk-off stance is time-bound to three to four weeks: the market is being asked to carry event risk through two binary catalysts — the October FOMC meeting and the midterm outcome — while real yields sit at a level that has historically compressed equity multiples. Hartnett has previously stated he expects growth to keep outperforming cyclicals into the midterms, which means market breadth could deteriorate before it improves: the index stays supported by a handful of mega-cap names while the internals become more fragile. The market looks healthier from 30,000 feet than it does on the ground. In that environment, his recommended "peak yield" nibbles include 30-year Treasuries, the Magnificent Seven, biotech, small caps, REITs, and Hong Kong property — a barbell that profits if long-end yields stabilize without requiring a broad equity rally.
Where the Fragility Actually Is
The survey's tail-risk ranking reads like a map of the leverage stacked beneath the AI rally. The biggest tail risk is "a disorderly rise in bond yields" at 31%, followed by "an AI bubble" at 23%, "a second wave of inflation" at 20%, and "a hyperscaler credit event" at 11%. When asked for the most likely source of a systemic credit event, 39% pointed to hyperscaler capital expenditure — surpassing government debt at 28%. Forty-five percent of managers expect an AI bubble to materialize in 2027.
Yet the positioning tells a different story. Fifty-seven percent of respondents do not anticipate cuts to AI capex, and roughly 60% expect double-digit corporate profit growth next year. Forty-seven percent expect above-trend growth and inflation — the highest level since February 2022. Bond pessimism has actually receded to its lowest level since May 2025, with the net percentage expecting long-term rates to rise falling to just 1%, down from 15% in September.
This is the second-order tension that matters. The market has priced a soft landing — 40% expect one, 50% a "no landing," and only 6% a hard landing — and it has priced continued AI spending. What it has not priced, in Hartnett's view, is the possibility that the bond market forces the AI build-out to be repriced through the cost of capital rather than through earnings. Hyperscaler capital expenditure is now the single largest candidate for a systemic credit event precisely because the equity market has treated it as inexhaustible. If the 10-year yield forces even a partial reassessment of that assumption, the repricing runs through semiconductor demand, utility build-outs, and the leveraged private credit that has quietly funded parts of the AI supply chain — not through the headline earnings of the hyperscalers themselves.
That is why the recommended contrarian trades are long bonds, REITs, and consumer discretionary stocks — the latter at its most underweight level since July 2008 — against short positions in equities, technology, banks, and emerging markets. "Buy humiliation and sell hubris," Hartnett wrote on October 1, meaning international equities, small-cap technology, and commodities over U.S. large caps, AI, and government bonds. He set the line at a 5% 10-year yield and warned that AI reaching 40% of the index would mark the top.
The Counter-Thesis: This Is a Cyclical Dip, Not a Regime Shift
The strongest case against Hartnett is that this is a cyclical fluctuation within a still-intact bull market, not a structural break. U.S. stocks are up 13.4% this year. The S&P 500, near 7,719, sits within striking distance of its 52-week high of 7,844. The VIX closed at 16.87 on October 14 — elevated from 15.27, but hardly at panic levels. Hard data has not shown a clear slowdown, and hyperscaler capex expectations for 2027 remain steeply rising. If AI-related capital spending continues to support real demand, there remains a rationale for yields to rise further without breaking equities — the same argument that carried the market through 2023 and 2024, when every rate scare turned out to be a buying opportunity.
That counter-thesis is credible, and it is the base case for the 76% of managers expecting the Fed to hold. But it rests on one assumption that Hartnett's framework directly challenges: that the bond market will tolerate further yield increases without a disorderly move. The evidence that the margin is thinning is specific and observable — the Korean brokerage breakdown, the French-German spread widening to 140 basis points, CCC credit underperforming, and the 10-year auction yield hitting levels not seen since 2000. A cyclical read requires those to remain contained. If they spread to U.S. duration assets, the cyclical thesis fails.
The falsifying signal is quantifiable: if the 10-year Treasury yield sustains above 5.5% for two consecutive weeks while the S&P 500 holds within 3% of its record high, the MOVE Index stays below 125, and high-yield spreads do not widen more than 50 basis points from their 2026 averages, Hartnett's risk-off call is wrong and the bull market is simply digesting a cyclical rate scare. Conversely, if the 10-year breaks 5.5% and the VIX pushes above 22 with financials leading the decline and credit spreads widening beyond those thresholds, the structural-repricing thesis is confirmed.
What to Watch Next
The next three to four weeks carry the two catalysts that will determine which path the market takes. First, the October 28 FOMC meeting: a hold keeps the passivity trade intact and leaves long-end yields to the market; a hike would be read as bullish under Hartnett's framework but would test risk assets in the near term. Second, the midterm election outcome, with the divided-government scenario priced as the least disruptive and a sweep of either kind carrying larger market implications. Third, the September core PCE print, projected at 0.32% month over month versus 0.25% in August — a hot read would reinforce the case for Fed action and push real yields higher, tightening the very constraint Hartnett sees as the market's binding limitation.
Split by time horizon, the picture is mixed. In the short term, sentiment and liquidity dominate: the drop in the sentiment indicator to 6.4, the rise in the VIX, and the defensive weekly flows all point to continued volatility. Over the medium term, fundamentals still support the market — 60% of managers expect double-digit 2027 profit growth, and AI capex remains intact, with global equity inflows on pace for a record year. Over the long term, the structural question is whether the bond bubble's end forces a permanent repricing of duration-heavy growth assets, or whether central-bank credibility proves sufficient to engineer a soft landing for both bonds and stocks.
The base case is continued choppiness through the Fed meeting and the midterms, with breadth deteriorating even as headline indices hold near records. The upside case is a Republican sweep or a divided government combined with a Fed that successfully caps long-end yields — either could add 5% to 10% to equities. The downside case is a disorderly move in the 10-year through 5.5% that forces the crowded long-semiconductor and long-Mag-7 trades to unwind simultaneously, with financials as the transmission channel.
Hartnett's warning, stripped to its essence, is that the market has spent two years betting that central banks would always rescue risk assets — and that bet is now priced into a bond market that has stopped believing it. The rally that survives the next month will be narrower, more expensive to finance, and far more dependent on earnings that have yet to prove they can justify the AI build-out. The bond bubble ended quietly. The question is whether the stock market notices before the midterms do.
Explore more exclusive insights at nextfin.ai.
