W.158Fund managers cut their tech overweight from 37% to 12%, BofA’s survey finds
Bank of America’s November fund manager survey, published on November 17, found a net 12% of managers overweight technology stocks, down from 37% in October. Reuters reported that as the largest monthly cut since June 2006. A record net 38% said companies are overinvesting, and 52% expect an AI hyperscaler to announce a capital spending cut in the next six months, up from 30% in October. The poll ran from November 6 to 12, before Nvidia’s results. Cash fell to 3.6% of portfolios from 3.9%, so managers did not report a move into cash.
The cut
Managers cut tech by 25 points in a month. The survey’s measure is the share of managers who say they are overweight a sector minus the share who say they are underweight. For technology it fell from net 37% in October to net 12% in November. Reuters describes that as the largest cut in a month since June 2006. The claim is about the size of the move. No report we read ranks the level against past months; a relay of Wallstreetcn’s account puts it about half a standard deviation below the long-term average.
October was a high. Notes on the October survey put that month’s net 37% as the highest since August 2020. The November reading is above February’s net 5% and below April’s net 14%. Managers remain overweight tech on balance. Fortune reports BofA’s view that the best entry points for tech come when investors are underweight, which they are not.
The poll closed before Nvidia reported. It was fielded from November 6 to 12. Nvidia reported its quarter after the close on November 17, the day the survey was published.
Two panel sizes are reported. Reuters gives 169 investors with $461 billion of assets. Investing.com gives 193 panelists with $526 billion. Neither explains the difference. The Wallstreetcn relay calls the 169 the global panel and gives a separate European panel of 94 with $241 billion.
Tech went from the year’s highest overweight to net 12% in one month
Net percentage of fund managers overweight technology stocks, by survey month in 2026, as press reports of each Bank of America survey gave it. Pink: November. Grey: earlier months. March, June, July and August are not shown because we did not find the reading in a report we could read; the bars are not evenly spaced in time.
What managers said about AI spending
A record share say companies are overinvesting. Investing.com and a relay of Wallstreetcn give it as a record net 38%. Reuters wrote a record 38% of investors, without “net”. The reading was net 33% in September and net 20% in November 2025, which Fortune reported as the first time since August 2005 that a net share said so. The question is about companies in general; some outlets added “in AI”.
More than half now expect a hyperscaler to cut. 52% expect an AI hyperscaler to announce a capital spending cut in the next six months, Reuters reported. In October 30% expected cuts and 57% expected none. From July to September the question asked about a cut this year, and the share expecting one fell from 28% to 14%. November’s horizon differs from the one asked from July to September, so the series is not like for like.
Consensus forecasts point the other way. Fortune cites FactSet estimates that capital spending by the hyperscalers Goldman Sachs tallies rises 58% next year to about $1.4 trillion. The survey measures what managers expect to be announced. It is not a forecast of the amount.
Managers want cash kept, not spent. A relay of Wallstreetcn reports that 42% want companies to improve balance sheets, 33% want capital returned and 16% want more capital spending, down from 34% in January. Fortune and Investing.com report the 16% as the lowest since May 2020.
52% expect a hyperscaler capex cut, up from 14% in September
Percentage of fund managers expecting an AI hyperscaler to cut capital spending, by survey month in 2026. Pink: November, when the question asked about an announcement in the next six months. Grey: earlier months, when it asked about a cut this year or in 2026; the October horizon is not stated in the notes we read. The label under each bar is the horizon. The change of horizon means November is not strictly comparable with the months before it.
The series
| Survey month, 2026 | Expect a cut | Expect none | Horizon asked |
|---|---|---|---|
| Jul | 28% | 61% | this year |
| Aug | 21% | 71% | this year |
| Sep | 14% | 79% | in 2026 |
| Oct | 30% | 57% | not stated |
| Nov | 52% | not reported | next six months |
Risks and crowding
Two AI answers take 65% of the tail-risk vote. 34% named an AI bubble as the biggest tail risk and 31% named a credit event at an AI hyperscaler. In October the two were 23% and 11%, 34% together, and the top answer was a disorderly rise in bond yields. The credit-event answer rose 20 points in a month.
Hyperscaler bonds lead a separate question on credit. Asked for the likeliest source of a credit event, 38% named AI hyperscaler bonds and 30% private credit. That 38% is a different figure from the net 38% who say companies are overinvesting.
Chip stocks are still the most crowded trade, by less. 39% named long global semiconductors, down from 45% in October and a record 82% in July. That is a share of respondents naming the trade. It is not a weighting.
Managers did not report raising cash. Cash fell to 3.6% of portfolios from 3.9%. A level at or below 4.0% is a contrarian sell signal under BofA’s rule, the Wallstreetcn relay says. Investing.com reports that managers trimmed their bond underweight by 13 points, and that 50% now expect a soft landing against 42% expecting no landing; October’s notes had 40% and 50%.
An AI bubble and a hyperscaler credit event are the top two tail risks
Percentage of respondents naming each as the biggest tail risk. Pink: the two AI answers in November. Grey: the other November answers reported. Faint grey: the same two AI answers in October, from notes on that survey. October readings for the other answers are not shown.
39% call chip stocks the most crowded trade, down from 82% in July
Percentage of respondents naming long global semiconductors as the most crowded trade, by survey month in 2026. Pink: November. Grey: earlier months. June is not shown because we did not find the reading. In February the leading answer was long gold, per reports at the time.
Series used in the charts
| Survey month, 2026 | Tech, net overweight | Semiconductors most crowded |
|---|---|---|
| Jan | 19% | not found |
| Feb | 5% | long gold led |
| Apr | 14% | 24% |
| May | 33% | 73% |
| Jul | not found | 82% |
| Aug | not found | 53% |
| Sep | 30% | 53% |
| Oct | 37% | 45% |
| Nov | 12% | 39% |
What this changes on inthesingularity.ai
- The hyperscaler capex funding pack works from company guidance. The survey adds an investor expectation to set beside it: 52% expect an announced cut within six months, while the FactSet consensus Fortune cites has spending up 58% next year.
- Lead: a small tracker of the survey’s AI questions by month: tech net overweight, net share saying companies overinvest, share expecting a hyperscaler cut, and the semiconductor crowding reading. The table above has the months we could source; March, June, July and August readings of the tech weight are missing.
- W.116 covered the Fed’s count of advanced-computing debt. The survey shows the same concern among investors: 31% name a hyperscaler credit event as the biggest tail risk, up from 11% in October.
- A date to watch: the December survey, the first fielded after Nvidia’s results.
Sources and method
| Item | Figure | Source |
|---|---|---|
| November survey | Tech allocation cut to net 12% overweight from 37%, the most in a month since June 2006; a record 38% say companies are overinvesting; 52% expect an AI hyperscaler to announce a capex cut in the next six months; 169 investors with $461 billion; 3% predict a recession | Reuters, via Investing.com, Nov 17 |
| November survey | Fielded November 6 to 12; 193 panelists with $526 billion; record net 38% overinvesting; capex-cut expectation 30% in October; long global semiconductors most crowded at 39%, from a July high of 82%; cash 3.6% from 3.9%; soft landing 50%, no landing 42%, hard landing 3%; credit event source: hyperscaler bonds 38%, private credit 30% | Investing.com, Nov 17 |
| November survey | Strategists led by Michael Hartnett; 16% want companies to increase capex, the lowest since May 2020; FactSet estimates have capex at the hyperscalers Goldman Sachs tallies up 58% next year to about $1.4 trillion; BofA says the best entry points for tech come when investors are underweight | Fortune, Nov 17 |
| November survey | Biggest tail risk: AI bubble 34%, AI hyperscaler credit event 31%, second wave of inflation 14%, disorderly rise in bond yields 13% | Seeking Alpha, via TradingView |
| November survey | Cash use: improve balance sheets 42%, return capital 33%, increase capex 16% (34% in January); cash at or below 4.0% is BofA’s contrarian sell signal; tech positioning about half a standard deviation below its long-term average; a European panel of 94 with $241 billion | Odaily relay of Wallstreetcn, via KuCoin, Nov 18 |
| Earlier surveys | October: tech net 37% overweight (September 30%); 30% expected AI capex cuts and 57% none; tail risks AI bubble 23%, hyperscaler credit event 11%; semiconductors most crowded at 45%, from 53% in September; no landing 50%, soft 40%, hard 6% | Atrani Capital, October survey notes |
| Earlier surveys | September: net 33% overinvesting, matching February’s record; 14% expected a hyperscaler capex cut in 2026 and 79% none | Trustnet, Sep 15 |
| Earlier surveys | August: 21% expected a capex cut this year and 71% none; semiconductors most crowded at 53%, from a record 82% in July | Trustnet, Aug 18 |
| Earlier surveys | July: 28% expected a capex cut this year and 61% none; semiconductors most crowded at 82% | XTB, Jul 14 |
| Earlier surveys | April and May: tech net 14% and 33% overweight; semiconductors most crowded at 24% and 73% | Trustnet, May 19 |
| Earlier surveys | January and February: tech net 19% and 5% overweight | Bloomberg, via The Wealth Advisor, Feb 17 |
| Earlier surveys | February: long gold the most crowded trade, cited by 50% | Benzinga, February 2026 |
| Earlier surveys | November 2025: net 20% said companies were overinvesting, the first time since August 2005 that a net share said so | Fortune, Nov 18, 2025 |
What we read. Press reports of the November survey and of earlier surveys. Bank of America does not publish the survey, and we did not read the report itself; every figure here is as an outlet gave it. Arithmetic. Tech cut: 37 − 12 = 25 points. Capex-cut expectation: 52 − 30 = 22 points on October. Overinvesting: 38 − 33 = 5 points on September. AI tail risks: 34 + 31 = 65; October 23 + 11 = 34. Capex as preferred use of cash: 34 − 16 = 18 points since January. Conflicts between sources. The panel is 169 investors with $461 billion in Reuters and 193 with $526 billion in Investing.com. The overinvesting reading is “net 38%” in two reports and 38% of investors, without “net”, in Reuters. February’s overinvesting record is given as about 35% in Bloomberg’s report at the time, without “net”, and as net 33% in September’s coverage, so we do not chart it. The capex-cut question is “an AI hyperscaler” in Reuters and “at least one” in Fortune. Limits. These are survey answers, not positions: a net overweight is a count of managers and says nothing about the size of holdings. The monthly series are assembled from different outlets’ reports and have gaps. The capex-cut question in November has a different horizon from the one asked from July to September; October’s is not stated. The poll closed five days before Nvidia’s results.
Wire reports are written within hours of the news, on a six-hour schedule. Each lists its sources and shows the arithmetic behind its figures. Not investment advice.