Key Takeaways
- • Jordi Visser argues October 2026 was a hedge-fund degrossing event within a bull market, citing Morgan Stanley prime brokerage data showing the largest October degrossing in at least ten years, while the S&P 500 fell only 93 basis points for the month.
- • Visser says breadth improved sharply, with the share of S&P 500 names above their 50-day moving average rising from 20% at the start of October to 61% at month-end, and nine of eleven sectors finishing higher.
- • The hyperscalers (Amazon, Microsoft, Google, Meta) have shifted from debt-free compounders to leveraged capex spenders trading like hedge fund positions; Visser recommends reducing hyperscaler weight and adding agentic-compute exposure.
- • Visser identifies the AI financing pipeline as the binding constraint, citing over $1 trillion raised for AI in 2026 across 16 companies and Oracle's best CDS week since September 2025 after it pledged to raise no more than $20 billion in additional financing through 2030.
- • Visser treats Anthropic's S-1, which disclosed $132 billion in cash, as the week's most underappreciated positive, and names specific calls including adding Micron and optical names and reducing Marvell if it acquires Lumentum.
Summary
Jordi Visser, a former portfolio manager who now publishes research through a subscription service, argues that October 2026 was a leverage-driven deleveraging inside a bull market rather than the start of an AI bubble unwinding. According to his review, the S&P 500 fell 60 basis points for the week ending October 30 and 93 basis points for the month, while nine of eleven sectors were up and 327 of 500 names rose at least 2%. He relies on Morgan Stanley prime brokerage data showing the largest degrossing in the month of October in at least ten years. Using ChatGPT for historical comparison, he finds that prior episodes with at least a 6% drawdown, including COVID, Trump 1.0, early 2018 and the 2016 shale unwind, each reached at least a 15% index drawdown, whereas the March 2021 GameStop unwind did not. He concludes October most resembles GameStop. Supporting this, the VIX touched above 20 before falling back and investment-grade CDX spreads stayed well below panic levels. Charlie McElligott's pair-trade analysis, which Visser cites, identified 124 pair trades in the unwind, 47 of which moved more than three sigma.
Visser acknowledges the technical picture looks bearish, including a head-and-shoulders formation in the S&P and Nasdaq, but argues the monthly charts look sound and the equal-weight index recovered. He frames the market structure as concentrated: roughly 150% of the S&P's total return since September 2025 came from a single 13-month window, and for NVIDIA four months out of thirty produced the entire 30% annualized move.
On the hyperscalers, Visser argues Amazon, Microsoft, Google and Meta now trade as high-beta positions. Google fell 12% relative to the S&P 500 in the week, its worst relative week since the financial crisis. He cites market-cap losses of roughly $250 billion for Meta, $500 billion for Google and $357 billion for Microsoft in 2026 drawdowns. He attributes the correlation with semiconductors to funds being underweight semis relative to the memory and inference opportunity, a mispricing he says has closed. He recommends reducing hyperscaler weight, increasing agentic-compute exposure, and holding longevity names such as Eli Lilly and energy. He characterizes this as a rotation regime rather than buy-and-hold, with multiple compression rather than credit risk as the main concern given the hyperscalers' balance sheets.
On financing, Visser argues investor absorption of AI debt, not compute demand, is the binding constraint. Citing a August 2026 Goldman Sachs Exchange discussion with Allison Nathan, he reports that financing needs are expected to peak in 2027 and that insurance company demand for long-duration debt fell roughly 50% due to concentration in the same issuers. Debt issuance declined each quarter as rates rose, with issuance shifting toward SPVs and private markets. Oracle was the week's pivotal development: its CDS had its best week since September 2025 after pledging to raise no more than $20 billion through 2030 and to reach positive free cash flow in the second half of 2028. Visser links this to Gavin Baker's view that cloud providers locked in capacity at lower prices, so renewals may accelerate cash flow. Visser flags the Treasury quarterly refunding around November 4, where Citi expects Bessent to cut long-bond sales, and the CCC market, with yields near 16%, which he expects to produce bankruptcies.
Visser views Anthropic's S-1, which showed $132 billion in cash, as a major positive. He estimated the company needed $200–250 billion, and a revenue scenario he ran produced a base case of $95–110 billion. He compares Anthropic's prior-year revenue of $1.4 billion with Google's $4.5 billion at IPO. Chamath Palihapitiya said on air the IPO should price in a $1.3–1.7 trillion range, and David Sacks argued the cash covers forward capex commitments through roughly 2030.
In memory and optical, Visser says SK Hynix beat expectations and Micron has contracts extending to 2030. He raised his Micron position and added optical names, while leaving Marvell untouched but planning to reduce it if the Lumentum acquisition proceeds. Corning's earnings, which management said reflected demand pressing against capacity, support his view that optical connectivity could expand tenfold per GPU. He also expects the 10-year yield peaked around October 7, 2026, and that the yield closed the week below its 50-day moving average for the first time in months.
Visser acknowledges the technical picture looks bearish, including a head-and-shoulders formation in the S&P and Nasdaq, but argues the monthly charts look sound and the equal-weight index recovered. He frames the market structure as concentrated: roughly 150% of the S&P's total return since September 2025 came from a single 13-month window, and for NVIDIA four months out of thirty produced the entire 30% annualized move.
On the hyperscalers, Visser argues Amazon, Microsoft, Google and Meta now trade as high-beta positions. Google fell 12% relative to the S&P 500 in the week, its worst relative week since the financial crisis. He cites market-cap losses of roughly $250 billion for Meta, $500 billion for Google and $357 billion for Microsoft in 2026 drawdowns. He attributes the correlation with semiconductors to funds being underweight semis relative to the memory and inference opportunity, a mispricing he says has closed. He recommends reducing hyperscaler weight, increasing agentic-compute exposure, and holding longevity names such as Eli Lilly and energy. He characterizes this as a rotation regime rather than buy-and-hold, with multiple compression rather than credit risk as the main concern given the hyperscalers' balance sheets.
On financing, Visser argues investor absorption of AI debt, not compute demand, is the binding constraint. Citing a August 2026 Goldman Sachs Exchange discussion with Allison Nathan, he reports that financing needs are expected to peak in 2027 and that insurance company demand for long-duration debt fell roughly 50% due to concentration in the same issuers. Debt issuance declined each quarter as rates rose, with issuance shifting toward SPVs and private markets. Oracle was the week's pivotal development: its CDS had its best week since September 2025 after pledging to raise no more than $20 billion through 2030 and to reach positive free cash flow in the second half of 2028. Visser links this to Gavin Baker's view that cloud providers locked in capacity at lower prices, so renewals may accelerate cash flow. Visser flags the Treasury quarterly refunding around November 4, where Citi expects Bessent to cut long-bond sales, and the CCC market, with yields near 16%, which he expects to produce bankruptcies.
Visser views Anthropic's S-1, which showed $132 billion in cash, as a major positive. He estimated the company needed $200–250 billion, and a revenue scenario he ran produced a base case of $95–110 billion. He compares Anthropic's prior-year revenue of $1.4 billion with Google's $4.5 billion at IPO. Chamath Palihapitiya said on air the IPO should price in a $1.3–1.7 trillion range, and David Sacks argued the cash covers forward capex commitments through roughly 2030.
In memory and optical, Visser says SK Hynix beat expectations and Micron has contracts extending to 2030. He raised his Micron position and added optical names, while leaving Marvell untouched but planning to reduce it if the Lumentum acquisition proceeds. Corning's earnings, which management said reflected demand pressing against capacity, support his view that optical connectivity could expand tenfold per GPU. He also expects the 10-year yield peaked around October 7, 2026, and that the yield closed the week below its 50-day moving average for the first time in months.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.