The traditional negative correlation between US Treasuries and tech stocks is being re-established in dramatic fashion.
Overnight, the 10-year US Treasury yield plunged more than 15 basis points in a single day, briefly breaking below the 5.00% mark to hit a one-month low, while the Nasdaq Composite tumbled 2.60% and the Philadelphia Semiconductor Index suffered a single-day drop of 6.72%.
Nomura cross-asset strategist Charlie McElligott noted in his latest report that the market is rapidly "repricing left-tail risk" 鈥?with the driving force having shifted from the previous inflation narrative to a systemic reassessment of AI-driven growth prospects. This shift means that US Treasury duration positions, which left holders "losing on both sides" over the past大半 year, are once again functioning as a hedge for equity portfolios.
McElligott made clear that this Treasury rally is no longer mere short covering, but rather the market beginning to characterize the AI sector selloff as a growth problem and positioning hedges accordingly. Goldman Sachs Delta One desk head Rich Privorotsky shares the same assessment, stating: "Bonds are rallying firmly alongside the AI selloff......the weakness in AI has repriced the capex cycle and is increasingly morphing into a growth shock."
Market Turn: From Right-Tail Fear to Left-Tail Pricing
Just two weeks ago, McElligott was describing an environment where the market's only concern was "missing the rally." The title of his latest report has now become "Rapid Repricing of the Left Tail."
Only 48 hours earlier, in another report, he judged that the brief Treasury short squeeze was still mechanical deleveraging and "does not yet constitute a narrative shift toward growth panic or recession expectations." But he left a caveat: "We could very well reach a 'growth panic' position in the near term." As it turned out, "near term" meant roughly 48 hours.
Thursday's market structure confirmed this shift. The S&P 500 closed down 1.22%, ending a three-day winning streak; the Nasdaq fell 2.60%; the Dow Jones Industrial Average edged up 0.02%, and the Russell 2000 gained 0.11%. Sector divergence was stark: consumer staples, healthcare and utilities led gains, while semiconductors, memory chips and optical communications led declines 鈥?a classic defensive rotation pattern.
Options Hedging Demand Surges Abruptly
McElligott noted that for weeks the market had seen almost exclusively demand for upside call options, a pattern that reversed on Thursday. Equity volatility, skew and volatility-of-volatility (i.e., "tail demand") all surged sharply.
The change in the QQQ implied volatility curve was particularly notable: front-month implied volatility jumped about 3 points from the prior day's close to around 24, above late-February levels, whereas just a month ago the front end was only in the 19-20 range. The relationship between Nasdaq spot and volatility also returned to textbook form: NDX fell, VXN rose, and the two reverted to the standard regression line.
Hedging flows recorded by Nomura's trading desk that day included: 1,300 SPX January 5900-strike puts, 18,150 VIX November 25-strike calls, and 2,000 SPX December 7400-strike puts.
The significance of the QQQ $715 level lies in this: once breached, market makers will enter negative gamma territory, and mechanical selling will become self-reinforcing. Nomura's estimates show that if the market falls another 1%, options market makers, leveraged ETFs and volatility-control funds combined would need to sell approximately $34 billion; a 3% drop would mean $134 billion; and a 5% drop would reach as high as $225 billion. On the upside, the same group would need to buy only $23.5 billion on a 1% rebound, and in a sharp rally scenario, volatility-control funds' selling would even leave them as net sellers overall.
Treasuries Regain Hedging Properties, Rates Market Prices In Cuts
In the rates market, the implications of this shift are even more profound. For most of 2026, the strategy of holding duration to hedge equity portfolios was nearly a "double loss" 鈥?stocks and bonds falling together rendered the traditional 60/40 allocation logic ineffective. On Thursday, however, Treasury yields fell to near one-month lows, and McElligott noted that dollar interest rate volatility also rose alongside risk-aversion sentiment, "US Treasuries are finally functioning as a hedge again."
US Treasury options buying observed by Nomura included: nearly 50,000 TYZ 108.5-strike calls, a cumulative 35,000 FVH 106-strike calls, and 23,000 and 17,000 TYZ 107.5 and 107-strike calls respectively.
The swaption volatility surface also confirmed this judgment: 1-month expiry, 1-year and 2-year tenor volatilities jumped 10.1 and 11.2 basis points in a single day respectively, while 10-year and longer tenors barely moved. McElligott noted that when the short end of the volatility surface moves first, the market is pricing the Fed's next move 鈥?which in a growth panic scenario means rising rate-cut expectations.
AI Narrative Reassessment: Chain Reaction from Anthropic to Meta
McElligott attributed the root cause of this selloff to a "rethinking of the AI trajectory" and laid out a detailed "indictment": Anthropic's annualized recurring revenue data disappointed the market and raised questions about its IPO viability; Gartner warned that more than 80% of emerging cloud service providers will face bankruptcy; multiple AI ecosystem-related IPOs met with cold receptions globally; Lam Research's (LRCX) guidance disappointed the market; and a broadly spreading expectation of "capex growth flattening" 鈥?even as Meta maintained its aggressive capex plans and raised its outlook.
The Meta case is particularly noteworthy. McElligott described hyperscalers' current predicament as "selling calls against themselves": Meta acknowledged in its 14C filing that it needs to raise "hundreds of billions of dollars" starting this autumn through issuing a new class of non-voting shares and substantial debt. This means hyperscalers are financing the AI arms race by selling their own upside to the public. The credit market is already pricing this risk: Meta's CDS spread widened from 80 basis points on July 29 to 120 basis points on October 22.
McElligott characterized this phenomenon as "non-linear" financing costs: when the world's most profitable AI spenders need to pay higher costs to borrow, the pressure on downstream emerging cloud service providers is multiplied 鈥?which also explains why emerging cloud service providers were the first link to collapse in this correction.
Growth Pillars Shake, Macro Narrative Faces Reassessment
McElligott noted that this selloff has gone beyond a single crowded trade. The two best-performing S&P 500 sectors this year 鈥?technology and energy 鈥?are precisely the two legs of what he calls the "core constraint of the AI cold war" barbell strategy, which has delivered a year-to-date return of 60% with a Sharpe ratio of 3. On Thursday, energy fell 1.7% and technology dropped 2.9%, both legs breaking simultaneously.
The deeper impact lies at the macro level. McElligott argues that the trickle-down effects of AI capital expenditure, combined with fiscal expansion by G10 governments, supported US nominal GDP remaining in a "hot" 6% to 10% range for most of this year, which in turn formed an important foundation for corporate earnings growth expectations. Once this pillar shakes, the earnings narrative, the overheating narrative and the "higher for longer" narrative will all come under simultaneous pressure.
McElligott wrote at the end of his report: "The previous assumption of a 'permanent capex trickle-down effect' no longer holds." The reflexive loop described by Goldman Sachs may provide some buffer 鈥?falling yields help ease hyperscalers' financing pressure, thereby supporting the AI trade. But this logic itself means the market is pricing in lower AI growth expectations, and with market makers in negative gamma below QQQ $715 and a 5% decline corresponding to $225 billion in mechanical selling pressure, the market may not be far from a true stress test.
Comments