Goldman Sachs: Market Breadth Hits Dot-Com Bubble Extremes, Momentum Factor Under Pressure from Both Sides
nashnova research
Goldman strategist Ben Snider warns that US market breadth has narrowed to dot-com-bubble extremes — the S&P 500 sits just 2% below its high, yet the median constituent stock is already down 15%, leaving the momentum factor exposed no matter which direction the market corrects.
The surface looks calm — what's happening underneath?
Goldman's cyclical-vs-defensive pair basket fell 6% over two weeks, tracking the firm's current activity indicator lower.
The 10-year Treasury yield dropped more than 20 basis points; US high-yield credit spreads widened over 20 bps, with pain concentrated in AI-linked issuers.
This means → capital is retreating from both "offensive assets" and "AI credit" simultaneously — stress is not confined to equities.
How severe is the AI sell-off?
Goldman's emerging cloud basket (GSXUNEOC) plunged 24.8% in a single week. Semiconductors fell for eight straight sessions — the longest losing streak since May 2006.
The high-momentum pair basket (GSPRHIMO) dropped 14.5%. Hedge-fund US long-short net leverage fell 3.2 percentage points in one week to 43.7%, the 0th percentile of the past five years and the largest weekly drop since "Liberation Day."
Net selling in information technology hit the highest level on record in Goldman's prime-brokerage data (since 2016).
In plain terms = hedge funds are unwinding AI and tech at extreme speed — leverage is at a five-year floor. This is not trimming; it is a stampede for the exit.
How long can AI capex keep boosting earnings?
Goldman estimates AI capital spending currently accounts for roughly half of S&P 500 EPS growth. But as capex growth slows and the depreciation bill — the annual cost of writing down equipment — arrives, that contribution will shrink each year.
By 2028, AI capex's contribution to earnings turns negative — depreciation flips from tailwind to drag.
This means → Goldman's 2028 EPS forecast of $460 is 7% below the bottom-up consensus of $492. The core disagreement is a single line item: depreciation.
What does Q3 earnings season tell us?
Among companies that have reported, 53% beat consensus by more than one standard deviation — above the 25-year average of 50%, but well below the 64% average of the past four quarters.
Earnings-revision breadth (the share of upward revisions minus downward) is still positive but narrowing alongside cyclicals.
This reflects a market shifting from "almost everyone beats" to "the beat rate is shrinking" — earnings growth has not collapsed, but the upside surprise is fading fast.
Why is the "buy momentum when growth slows" playbook breaking?
Historically, when real growth is positive but decelerating, the S&P 500's median quarterly return is 3% and momentum — buying recent winners, selling recent losers — is the best-performing factor, with a median long-short quarterly alpha of 1.8 percentage points.
Snider explicitly warns that today's extreme breadth compression creates a major exception: the S&P 500 is just 2% from its high, yet the median constituent is already down 15%.
In plain terms = the index looks fine only because a handful of mega-caps are holding it up; most stocks have already fallen sharply. A momentum strategy right now piles chips into exactly those few crowded, fragile names.
What comes next?
Snider offers two response baskets: one for breadth recovering upward (laggards catch up) and one for breadth converging downward (leaders catch down).
Under either scenario, today's highly concentrated momentum positioning faces pressure.
This means → the key checkpoint is whether the transmission of AI-financing fears from credit markets into equity-strategy positioning is complete — if it is not, the momentum unwind may have further to run.
市场有风险，内容仅供研究参考，不构成投资建议。