Fund Managers Cut Tech Allocations at Fastest Pace Since 2006
BofA's November survey shows tech overweight falling from 37% to 12%, the sharpest cut since June 2006, as a record 38% of managers call AI investment excessive.

Why it matters
- Net overweight in technology fell from 37% to 12% in November, the largest one-month drop since June 2006.
- 52% of managers expect an AI hyperscaler capex cut within six months, up from 30% in October.
- Average cash holdings of 3.6% are described by BofA as extremely low, so the survey is not a buy signal.
- S&P 500 fell 1.03% to 7,340.53 on the survey release date, its fourth straight lower close.
- NVIDIA made up 14.65% of XLK and 7.6% of QQQ as of June 30.
The net share of global fund managers overweight technology fell to 12% in November from 37% in October, a 25-point drop that Bank of America calls the largest one-month decline since June 2006. The figure comes from Bank of America's November Global Fund Manager Survey, released November 17, 2026, and it marks a sharp reversal in a sector that still anchors the two largest technology ETFs traded in the United States.
The scale of the survey matters for how much weight the number can carry. Between 169 and 193 managers, overseeing assets between $461 billion and $526 billion, answered the global poll between November 6 and November 12. The survey measures how managers say they are allocated rather than what they actually traded, and technology remains a net overweight even after the cut. The direction of travel, not the level, is what changed: managers are reducing exposure at a pace not seen in two decades.
Markets fell on the day the survey landed. The S&P 500 dropped 1.03% to 7,340.53, its fourth straight lower close, down from 7,619.38 on November 11. The State Street Technology Select Sector SPDR ETF (NYSEARCA:XLK) closed lower at $163.26. The Invesco QQQ Trust (NASDAQ:QQQ) fell 0.73% to $656.95. Salesforce (NYSE:CRM) dropped 9.45% to $194.52, the sharpest single-name move among the names cited alongside the release.
Both funds recovered after hours once NVIDIA (NASDAQ:NVDA) reported earnings, with XLK at $166.78 and QQQ at $665.93. After-hours quotes can change before the next open, and one earnings report does not settle the allocation question the survey raised.
Why are managers turning against AI spending?
The core of the shift is capital expenditure. According to the survey, 52% of managers now expect at least one AI hyperscaler to announce a capex cut within six months, up from 30% in October. No hyperscaler has confirmed any such cut; the forecast is investor expectation, not announced policy. Even as an expectation, it moves markets, because hyperscaler capex is revenue for chipmakers and the rest of the AI supply chain. When investors push the biggest buyers of AI infrastructure to spend less, they threaten the sales of the stocks that carry the most weight in XLK and QQQ.
A record net 38% of managers say companies are overinvesting in AI. Only 16% want companies to raise capital spending, the lowest share since May 2020. Managers named an AI bubble as their biggest tail risk — ranked ahead of every other risk in the survey.
The survey also picked up a new concern in the credit market. In a separate question, 38% of respondents named AI hyperscaler bonds as the most likely source of a credit event, ahead of private credit at 30%. AI buildouts are increasingly paid for with debt. If bond buyers demand higher yields, the spending that supports chip revenue becomes harder to finance. The result, as the survey data lays it out, is that stock investors and bond investors now share the same risk exposure concentrated in the same small group of hyperscalers.
Is the selling a bottom signal?
Bank of America's own strategists say no. Average cash holdings stand at 3.6%, which the bank describes as extremely low. Almost no manager expects a recession, and the average forecast for the S&P 500 at the end of 2027 is 7,926 — above the November 17 close of 7,340.53. Managers cut tech yet stayed fully invested and optimistic.
That combination has a historical read. Bottoms, by the pattern the survey team tracks, usually come after managers build up cash and turn gloomy. Here they did neither, which means they retain room to sell more. The team led by strategist Michael Hartnett reads the November results as investors telling companies to pause AI capex, and it reads the data as a signal that the AI boom is ending. The team says the survey does not yet show enough fear to serve as a buy signal, and it remains tactically risk-off.
Interest rates add a second layer of pressure. The Federal Reserve meets December 8-9, and a rate hike is widely expected. Higher rates hurt highly valued growth stocks most, and the largest positions in XLK and QQQ are exactly those stocks.
How concentrated is the exposure?
Index funds hold a large share of this risk. As of June 30, NVIDIA alone made up 14.65% of XLK and 7.6% of QQQ. XLK's next two holdings weighed 12.85% and 8.38%. A fund in which one chipmaker approaches 15% of assets transmits any AI capex decision directly to millions of passive investors.
The recent performance numbers show the pressure already at work. XLK fell 11.02% over the past month and QQQ fell 7.45%. XLK remains up 13.8% for the year, a buffer if the AI trade weakens further, but a buffer shrinks with each leg down.
What could change the outlook for XLK?
Two events would suggest the pressure is letting up, according to the survey's own indicators. First, if the December survey shows cash holdings rising well above 3.6%, managers would be building the defensive position that historically precedes a bottom. Second, if XLK holds above its November 17 close of $163.26 after the December 8-9 Fed meeting, the rate-driven leg of the selling may have run its course.
The structural alternative, as the survey coverage notes, is diversification away from the capex cycle itself. An equal-weight S&P 500 fund spreads exposure so that no single chip supplier drives results. Companies that use AI to cut costs carry different risks from those building AI infrastructure, and the November survey draws exactly that distinction — the market is starting to price the builders and the users differently.
The stakes extend past one ETF. Hyperscaler capital spending is the revenue line for chipmakers, the issuance pipeline for the corporate bond market, and, through index concentration, the largest single exposure in passive equity portfolios. A survey of 169 to 193 managers will not cut a single capex budget, but 52% of them now expect someone else to — and the December 8-9 Fed meeting, four weeks of hyperscaler commentary, and the next monthly survey will test whether that expectation becomes an announcement.
Source: 247wallst.com
More from Rebecca Stone
Show full bio
Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.
754 articles
Related articles
- Hedge Funds Dumped Tech at Near-Record Pace, Then Chips Jumped 7%
- Goldman Sachs Sees $1.2 Trillion Big Tech AI Spending Wave by 2027
- Chicago Fed's Goolsbee Calls AI Data Center Boom 'Old-Fashioned Overheating'
- AI Data Center Stocks Crater as Capital and Politics Tighten
- Nvidia Is Just 17% of SMH — Yet the Chip ETF Jumped Nearly as Much