Semiconductors suffered a historic rout, SOXX fell 29.7% in one month (a 1-in-6,000 event), dragging the Nasdaq-100 down 13.4% as the market repriced AI capex from growth story to capital black hole.
▸
The Fed's pause masks a private tightening, Hyperscalers burned through 95.5% of operating cash flow on AI infrastructure ($691B TTM), collapsing free cash flow 83.7% YoY to $33B while the Fed holds rates at 3.75-4.00%.
▸
Crypto infrastructure trust fractured, Binance's $6.3B hack triggered $4.7B in net withdrawals and $1.28B in spot Bitcoin ETF outflows; Fear & Greed hit "Extreme Fear" (28) for the first time since August.
The week that was
The story tying this week together isn't the semiconductor crash, it's the capital wave behind it. Five hyperscalers (Microsoft, Google, Amazon, Meta, Oracle) deployed $691 billion on capital projects in the trailing twelve months through September, up 93.5% year-over-year while revenue grew only 19.1%. That capex now consumes 35.8% of every revenue dollar and 95.5% of operating cash flow.
Aggregate free cash flow evaporated from $201 billion to $33 billion, an 83.7% collapse. The market's verdict arrived in semiconductors first: SOXX down 29.7% in a month (z-score -3.6), with Intel and Marvell each shedding roughly half their value in four weeks. Friday's bounce, AMD +11%, Intel +7.7%, TSMC +6.3%, recovered only a fraction; the index remains 37% below its June peak.
That same capital vacuum appears in the bond market. The 10-year Treasury yield held at 4.84%, the 2-year at 4.46%, and the 10-year/3-month spread at 0.67 percentage points, still inverted but steepening. The VIX jumped to 25.64, its highest since April 2025. High-yield spreads widened to 3.48 percentage points as non-AI corporates face capital scarcity.
The dollar index firmed to 102.3, and the broad trade-weighted dollar sat at 121.7, close to the 122 level that historically triggers emerging-market funding stress. Mortgage rates at 7.18% continue to freeze housing: new-home sales run 5.4% below the 12-month average, and the starts-to-permits ratio has compressed to 87.6%.
Oil told two stories at once. Brent futures fell to $98.32, down 16% from the October 30 high of $117.64, as midterm election results raised hopes of an Iran peace deal. But dated Brent spot, the physical market, still traded at $107.03 on November 3, and the dual chokepoint crisis persists: Iran's exclusion zone in the Strait of Hormuz and Houthi control of the Bab al-Mandeb remain active.
Saudi Arabia's East-West Pipeline moves up to 5 million barrels per day around Hormuz but has not fully capped the risk premium. U.S. crude stocks sit at 409.8 million barrels with refineries running at 94.5% utilization. Natural gas storage rose 77 bcf to 3,801 bcf, easing the previously tight gas market.
Crypto's week was defined by the Binance aftermath. Bitcoin traded at $75,739 (down 8.3% on the month but up 16.8% on the quarter), Ether at $2,231. Stablecoin supply contracted to $292 billion, down 4.65% in 30 days, down 3.35% year-over-year. Spot Bitcoin ETFs saw five consecutive days of net outflows totaling $1.28 billion through November 6. The Fear & Greed index at 28 signals extreme fear; a month ago it was 59.
Gold rallied 4.5% on the week to $4,075, defying the stronger dollar and higher real yields as sovereign buyers continue accumulating. The Nikkei plunged 18% in a month (z-score -3.9), the Brazilian Bovespa 17.5% (z-score -3.8), reflecting a global risk-off wave that hit export-heavy indices hardest. Defensive sectors outperformed: consumer staples, healthcare, financials, and utilities all hit new 52-week highs even as the Nasdaq-100 corrected 13.4%.
Earnings season delivered beats with asterisks. McDonald's beat by 6.2% on global comp sales of +1.5%; Novo Nordisk beat by 9.9%; ConocoPhillips beat by 6.9% and raised its dividend 11%; Sony beat by 26% on record quarterly sales.
But Uber missed GAAP EPS by 73% due to equity-investment revaluation losses, and ARM missed EPS by 32% with royalty revenue below expectations. Qualcomm's slight EPS miss alongside a revenue beat pointed to margin pressure from memory costs and reduced Apple share. Berkshire Hathaway reported a 28% drop in quarterly net income as investment gains shrank and insurance profits fell.
The Federal Reserve held rates at 3.75-4.00% on October 28 (10-2 vote, Governors Cook and Hammack dissenting for a hike). The statement upgraded economic activity from "solid pace" to "strong pace." Chair Warsh emphasized inflation has been elevated for nearly six years. Market pricing for a December 25-basis-point hike stands at 78-80% as of November 7. The pivot point is Tuesday's CPI: core month-over-month forecast at 0.2% (prior 0.3%). A print ≥0.3% makes a December hike near-certain; ≤0.2% drops probability below 50%.
The big question this week
Can the AI capex wave sustain itself without breaking the rest of the economy?
This is functioning as a private-sector tightening regime the Fed neither sees nor offsets. The NFCI at -0.459 confirms aggregate corporate investment is contracting outside the megacap AI complex. High-yield spreads at 3.48% are widening as non-AI corporates face capital scarcity. Housing shows the bite: 30-year mortgages at 7.18% plus capital scarcity have pushed the starts-to-permits ratio to 87.6% and new home sales 5.4% below the 12-month average.
Emerging markets take the next hit: the broad trade-weighted dollar at 121.7 and USD/JPY at 158.4 signal a global dollar shortage the Fed's domestic mandate ignores. Gold at $4,075 and oil near $100 despite a soaring dollar confirm the correlation breakdown, but the driver is capital flight to U.S. AI assets, not inflation hedging alone.
To fund the build-out, hyperscalers issued $230 billion in investment-grade bonds in the first nine months of 2026, more than double all of 2025, raising debt's share of capex funding from 9% to 32%. Amazon alone burned $42.4 billion in trailing FCF; Alphabet posted a second straight quarter of negative FCF (-$9.1 billion). Microsoft guides FY2027 capex near $170 billion. Hyperscaler capex is only ~25% chips; the rest is data centers, power, cooling, networking.
If the agent-AI wave (GPT-5.5, Gemini 2.5 Pro, Claude Computer Use, Grok 3) replaces meaningful white-collar labor within 12-18 months, the productivity payoff validates the spend and semis bottom. If adoption lags, law firms and consultancies are still hiring strongly per this week's frontier data, the $691 billion capex becomes a stranded asset class, HY spreads blow past 4%, and the Fed is forced to cut into sticky core PCE (3.12%) to relieve the private tightening it didn't create.
Watch: Hyperscaler Q4 capex guidance (due January 2027) and January SOXX earnings. If capex growth decelerates below 30% year-over-year and semis stabilize, the crowding-out eases. If capex re-accelerates while semis make new lows, the credit event in non-AI corporates arrives by Q2 2027.
The week ahead
Tuesday, November 10, U.S. CPI (October data), 8:30 AM ET Core CPI m/m forecast 0.2% (prior 0.3%); headline CPI m/m forecast 0.3% (prior 0.6%). This is the single most important print before the December 8-9 FOMC. A core print ≥0.3% locks in a December hike; ≤0.2% makes it a coin flip. The 2-year yield (currently 4.46%) will move first, it's the Fed-policy proxy.
Wednesday, November 12, UK GDP m/m, 2:00 AM ET Forecast 0.0% (prior +0.1%). The Bank of England hiked to 4% last week for the first time since 2023; a flat print reinforces the stagflation bind that could force the BOE to pause despite above-target CPI.
Thursday, November 13, U.S. PPI (October data), 8:30 AM ET Core PPI m/m forecast 0.3% (prior 0.7%); headline PPI m/m forecast 0.4% (prior 1.3%). PPI leads CPI at the factory gate, a soft print here would reinforce the disinflation narrative and pull forward rate-cut expectations for 2027.
Earnings to watch (pre-market Monday): KKR ($88B), Constellation Energy ($85B), Barnes Group ($66B). After-hours Monday: Occidental Petroleum ($56B), Take-Two ($40B). Wednesday: Cisco ($210B). Thursday: Walmart ($650B), the consumer bellwether.
Worth knowing: Drawdowns and staying in
The lesson: Drawdowns are the price of admission for long-term returns. The S&P 500 has spent roughly 75% of all trading days below its all-time high. Since 1928, the median bear-market decline is -33% and takes 13 months to bottom, but the recovery takes 27 months on average.
The worst part? The best days cluster next to the worst. Missing the 10 best days since 1990 cuts your annualized return from ~9.5% to ~5.5%.
Missing the 20 best days drops it to ~3%. Those best days almost always occur during a drawdown, not after it ends.
Applied to this week: SOXX is down 37% from its June high. The Nasdaq-100 is down 13.6%. If you sold at the lows Friday, you locked in a historic loss and missed the 5-11% bounce the same session. The investor who stays in doesn't need to time the bottom, they just need to avoid turning a paper loss into a realized one.
The mechanism is simple: time in market > timing the market, because the distribution of returns is fat-tailed and negatively skewed. Your 401(k) contribution this paycheck buys more shares at lower prices. That is the only "strategy" that has worked across every cycle.
Sharper edge: The drawdown duration matters more than depth for compounding. A 30% drop that recovers in 6 months costs you less than a 15% grind that lasts 3 years, the latter steals time, and time is the only variable you can't recover. The current semis drawdown is deep (-37%) but young (5 months). The 2000-2002 tech bust took 31 months to bottom.
The 2008 financial crisis took 17 months. If this is a cyclical capex digestion (like 2018 semis -30%, recovered in 10 months), the bounce is tradable.
If it's structural (capex stranded, FCF permanently impaired), the drawdown extends. The tell: watch hyperscaler FCF inflection. When aggregate FCF stops declining, the drawdown clock starts ticking toward recovery.
Tomorrow's setup: Tuesday's U.S. core CPI m/m (forecast 0.2%, prior 0.3%). If the print comes in above forecast, which way does the 2-year yield move, and why? We'll walk through the transmission chain in tomorrow's edition.
Not financial advice. This brief is for informational and educational purposes only and does not constitute investment advice or a recommendation to buy or sell any security.
Data: macro indicators per FRED® (Federal Reserve Bank of St. Louis); not endorsed or certified by the Federal Reserve Bank of St. Louis. Energy data per the U.S. Energy Information Administration (EIA). Auction data per U.S. Treasury Fiscal Data. Filings per SEC EDGAR. Market prices per Yahoo Finance. Earnings calendar per Financial Modeling Prep. Crypto data per Binance and CoinGecko. Nikkei plunges 18% in a month as global tech unwind hits Japan's semiconductor-heavy index The Nikkei 225 fell to 57,435 on November 6, down 18.0% in one month (z-score -3.9) and 12.5% in three months (z -2.1), erasing most of its 2026 gains. The selloff is fundamentally a passive spillover from the global semiconductor rout: Japan's index carries outsized weight in chip-equipment makers (Tokyo Electron, Advantest, Disco) that trade as beta to Nvidia and the AI capex cycle. A weaker yen (USD/JPY 158.4) failed to cushion exporters because the driver is not currency but synchronized de-risking, VIX at 25.6, 20y+ UST ETF +6.2% monthly, HY OAS at 348 bps, signaling growth fears that hit cyclical, leveraged equity exposure hardest. The index is still up 13% year-on-year, so the move is a violent correction from the June peak, not a bear-market breakdown. Continuation requires further semiconductor earnings cuts or credit-event widening; reversal needs SOXX stabilization, a Fed pause signal, or BOJ liquidity reassurance that stops the forced unwind of leveraged long-Nikkei/short-JPY carry positions.