Tech stocks see the sharpest underweighting in 20 years! BofA survey reveals the current state of the "AI trade": positioning has cooled sharply, and the panic needed to buy the dip is still insufficient.
A Bank of America survey shows that investors' optimism about the global AI boom is fading, and fund managers are sharply cutting their investments in tech stocks by the largest margin in two decades. The survey, conducted from November 6 to 12, found that the proportion of fund managers overweight tech stocks fell from 37% in October to 12%, the largest drop since June 2006.
Title context: Tech stocks see the sharpest underweighting in 20 years! BofA survey reveals the current state of the "AI trade": positioning has cooled sharply, and the panic needed to buy the dip is still insufficient.
Text:
The latest survey report from Wall Street financial giant Bank of America, targeting the world's top fund managers, shows that the strong rebound in the core of global technology stocksnamely AI computing-power theme stocks and the broad semiconductor sectorhas not yet translated into a sustained, expanding recovery in risk appetite for global technology stocks. The BofA survey shows that institutional investors' optimism toward the global artificial intelligence investment boom has been fading sharply in recent years, with fund managers cutting technology stock exposure by the largest margin in nearly two decades. This BofA survey had 169 participants managing a combined US$461 billion in assets.
The latest BofA fund manager survey shows that the stock market frenzy around AI is undergoing a significant cooling in allocation, but the broad pessimism needed to support contrarian buying has not yet formed. From November 6 to 12, the net percentage of fund managers overweight technology stocks fell from 37% to 12%, a 25-percentage-point drop in a single month, the largest decline since June 2006; at the same time, the net percentage believing companies are overinvesting reached a record 38%, and 52% of respondents expect an AI hyperscaler to announce capital expenditure cuts in the next six months, indicating that the market has begun to re-examine the match between computing-power expansion and end-market revenue, financing capacity, and return on investment.
However, the BofA survey also shows that fund managers globally remain net overweight technology stocks, while the average cash allocation of funds instead fell to 3.6%, and only about 3% of respondents expect a recessionenough to show that capital is mostly being reallocated between sectors and regions rather than broadly retreating into cash assets. This is precisely the important basis for BofA Chief Strategist Michael Hartnett, known as "Wall Street's most accurate strategist," to advocate maintaining a defensive stance in the coming monthsreduced technology stock crowding does not automatically mean that the earnings expectation downgrade risk and valuation collapse pressure driven by data center construction delay risk after the Democratic Party's midterm election victory have been fully priced in.
Compared with Morgan Stanley and Goldman Sachs, which are bullish on medium-term equity returns, BofA is more focused on whether current positioning can withstand new expectation gaps; for the global AI industry chain that is crucial to the technology stock rally, the next stage of investment divergence will depend more on who can still convert orders into sustainable profits and cash flow in an environment of slowing growth in infrastructure-related hardware procurement tied to AI data center construction and tighter financing constraints.
The AI frenzy recedes, fund managers reallocate risk exposure! Investors cut technology exposure by the largest margin since 2006
Overall, BofA's survey report targeting global fund managers shows that institutional investors' optimism about the global artificial intelligence boom is fading, and fund managers have cut technology stock exposure by the largest margin in nearly two decades. The survey, conducted from November 6 to 12, found that the net percentage of fund managers overweight technology stocks fell from 37% in October to 12%, the largest drop since June 2006.
Michael Hartnett wrote in the report that investors also sent a "strong" signal: companies should pause artificial intelligence spending. The net percentage of respondents believing companies are overinvesting reached a record 38%, and another 52% of respondents expect an AI hyperscaler to announce capital expenditure cuts in the next six months.
Hartnett's latest commentary on the AI trade theme wrote: "The frenzy around the economic boom party and AI appears to be over."
As shown in the chart above, technology stocks are heading for their worst November performance since 2008investors are withdrawing funds from AI-related hot stocks. Note: 2026 data as of the November 16 close.
A new round of concerns about the returns on massive AI spending has hit global financial markets. The technology-heavy Nasdaq 100 Index has fallen more than 4% this month and, based on the current trend, may record its worst November performance since 2008. A semiconductor-related stock index has fallen 27% since early October.
BofA's Hartnett advises maintaining a risk-averse stance in the coming months because the survey "has not yet shown enough investor 'fear' to support buying." He said that based on the historical experience of this survey, the best entry point tends to appear after fund managers further cut technology stock exposure and turn underweight.
However, investors are less concerned about risk spreading to the broader economy. Only about 3% of survey participants expect a recession in the next year, while cash holdings as a share of total assets fell to 3.6%, the lowest since August.
Other market strategists are also bullish on equities. This week, the Morgan Stanley team said the AI spending boom will make 2027 "a year to embrace risk"; Goldman Sachs' forecasting team expects the S&P 500 to rise 17% over the next 12 months to 8,700 points.
In the BofA survey, about 36% of fund managers expect international stocks to be the best-performing asset in 2027, followed by U.S. stocks and gold at 24% and 17%, respectively. BofA's survey notes show that this survey had 169 participants managing a combined US$461 billion in assets.
Rebound sold off again, market begins to simultaneously cut AI growth expectations and valuation premiums
The strong rebound in AI computing-power stocks has not yet translated into a sustained recovery in risk appetite. From November 9 to 13, the Philadelphia Semiconductor Index and the Nasdaq 100 Index rose about 8.05% and 3.53%, respectively; but on November 16, they fell again by 5.28% and 2.67%. The Philadelphia Semiconductor Index closed at 9,610.65 points, about 34.3% below its highest close of 14,634.72 points on June 22, still deep in technical bear market territory.
These latest market data highlight that a phased counterattack can occur during a relatively deep correction, and the rebound itself is not enough to prove that forward earnings expectations have stabilized.
Nvidia announced on November 12 that Jensen Huang would step down, with Chief Financial Officer Colette Kress serving as interim chief executive officer, further raising market attention to strategic execution and continuity of customer relationships. Amid this round of uncertainty, Goldman Sachs cut its target price from US$300 to US$220, Citi from US$285 to US$225, and UBS from US$300 to US$240, declines of about 26.7%, 21.1%, and 20%, respectively, but all retained "Buy" ratings.
Goldman Sachs believes the backlog still supports operating performance over the next 1218 months, and uncertainty has a greater impact on valuation multiples; UBS cut its 2028 revenue and earnings per share forecasts by about 10% due to potential export restrictions and data center delays. However, it should be noted that this round of market repricing has simultaneously involved forward earnings and the valuation investors are willing to pay, and cannot simply be understood as a comprehensive disappearance of near-term orders.
Asian markets amplified this pressure further. On November 16, Taiwan's weighted index, South Korea's KOSPI, and Japan's Nikkei 225 closed down 6.03%, 3.59%, and 2.42%, respectively; on the 17th, Taiwan and South Korea fell another 2.31% and 2.59%, for cumulative two-day declines of about 8.20% and 6.09%, respectively. At the industry chain level, changes in GPU procurement expectations affect demand for advanced processes, packaging, HBM, and high-speed connectivity, so holding stocks in different regions and different sub-sectors may still jointly expose investors to the same AI capital expenditure cycle. Leadership transitions, financing constraints, and project deployment expectations acting together can cause ostensibly diversified portfolios to suffer synchronous drawdowns during stress periods.
Technology cut but no retreat to cash, BofA warns gap remains between "easing crowding" and "forming a buying point"
The most noteworthy aspect of the BofA survey is that investors' attitude toward AI has turned cautious, but overall fund manager positioning indicators have not yet shown a sufficiently defensive investment stance.
From November 6 to 12, the net percentage of fund managers overweight technology stocks fell from 37% to 12%, a drop of 25 percentage points, the largest decline since June 2006; this is a change in respondents' allocation stance, not a 25-percentage-point decline in technology stocks' share of portfolio assets.
At the same time, the net percentage believing companies are overinvesting reached a record 38%, and 52% of respondents expect an AI hyperscaler to announce capital expenditure cuts in the next six months; according to a supplementary disclosure by the Financial Times, the latter figure was only 30% last month, meaning confidence in continued capital expenditure expansion has changed markedly.
However, the average cash allocation of fund managers instead fell to 3.6%, and only about 3% of respondents expect a recession in the next year, indicating that the adjustment is closer to risk reallocation between sectors and regions. BofA's November 12 "The Flow Show" already noted: if the cash ratio falls below 3.7%, the market still lacks sufficient fear; fund flow data tracked in the same report showed that global equity funds still had net inflows of US$98.3 billion that week, and technology funds had net inflows of US$8.9 billion. Although fund flows and survey positioning are different samples, together they show that cautious market commentary/tone has not yet fully translated into withdrawals or underweighting.
Hartnett pointed out that if the fund manager survey report shows institutional investors' expectations for next year's economic "Boom" remain high, and fund managers' average cash allocation ratio falls below 3.7%, this means market panic is not yet sufficient, capital remains crowded in the market, and the correction in U.S. and even global technology stocks is not over. The team led by BofA strategist Hartnett is waiting precisely for market risk-taking willingness to contract further to the point of a major bottom-fishing opportunity, rather than confirming a buying window merely because technology stocks have fallen a lot.
The latest model adjustment by well-known semiconductor industry chain research firm SemiAnalysis provides an industry-level explanation for the capital expenditure doubts in the BofA survey: slowing growth in new annualized revenue for frontier model companies, combined with tighter debt and equity financing, has weakened the ability of some projects to lock in land, power, and machine rooms in advance.
SemiAnalysis cut its 2028 global AI accelerator spending forecast by 28% to US$758 billion and its 2030 forecast by 40% to US$923 billion; but the new model still expects spending to continue growing from US$392 billion in 2026. It should be noted that what has been revised down here is the previously expected expansion path, and the scope is concentrated on accelerators, and cannot be expanded to all data center capital expenditure.
The institution also expects that of the U.S. data center capacity originally scheduled to come online between the second half of 2027 and the first half of 2028, 5.8 gigawatts will be delayed, but emphasizes that revenue growth and the financing environment are the more important reasons for the adjustment. The institution's analysts said that, broadly consistent with the forecast model, investors are demanding that future procurement scale be re-matched with paid demand, financing capacity, and actual delivery progress.
From the perspective of hardware economics, procurement amount is determined jointly by delivery volume and average selling price, while actual computing output also depends on performance, software efficiency, utilization, and power supply conditions. SemiAnalysis cut its five-year cumulative logic wafer shipment area forecast by 17%, its 2028 HBM demand forecast by about 30%, and lowered HBM price and some GPU selling price assumptions, showing that the adjustment involves both demand scale and supplier pricing.
SemiAnalysis said Nvidia accounts for 66% of the industry's wafer area forecast reduction, reflecting its greater exposure to high-priced GPUs and new cloud customers under financing pressure; by contrast, Broadcom's ASIC business five-year cumulative revenue forecast was cut by 11%, and Amazon Trainium's 20272028 shipment forecast was unchanged. From this, the competitive focus is expanding to unit effective computing cost, energy efficiency, customer financial strength, and data center deliverability: slower capital expenditure growth will lower revenue expectations for some hardware suppliers, while lower computing costs may also improve returns for cloud platforms and application companies. This is precisely the direction in which the capital allocation divergence revealed by the BofA survey needs to be further identified.
In the BofA survey, 52% of respondents expect cloud giants to cut capital expenditure; SemiAnalysis's forecast cuts further reveal how this concern may transmit through the industry chain: slower procurement expansion and lower hardware prices will lower revenue expectations for some chip suppliers while creating conditions for computing-power users to reduce costs. The BofA strategist team said investors' AI allocation strategies therefore need to further distinguish which suppliers can defend share and profit margins, and which cloud platforms and application companies can convert cost declines into actual operating returns.
Related Articles

Fed's Hammack calls for clear rules on Treasury market interventions, raising the bar for entry and establishing a clear exit mechanism.

Fed Governor Waller Strongly Backs Payment Innovation: Supports Blockchain and Asset Tokenization Development, Will Explore Integration of New Technologies with AI

Can consumer resilience prop up the U.S. stock market rebound? U.S. retail data beat expectations, leading the market to stand firm against the most intense tech-stock selloff in 20 years.
Fed's Hammack calls for clear rules on Treasury market interventions, raising the bar for entry and establishing a clear exit mechanism.

Fed Governor Waller Strongly Backs Payment Innovation: Supports Blockchain and Asset Tokenization Development, Will Explore Integration of New Technologies with AI

Can consumer resilience prop up the U.S. stock market rebound? U.S. retail data beat expectations, leading the market to stand firm against the most intense tech-stock selloff in 20 years.






