Bank of America rings the "bull market alarm": AI capital expenditures may face liquidation, suggesting defensive portfolio adjustments for funds.

As the chip sector experiences intense volatility and Micron and Qualcomm briefly boost market confidence with strong performance signals, Sebastian Raedler, head of European stock strategy at Bank of America, issued a completely different warning: if artificial intelligence trading begins to unravel, investors should decisively shift to defensive industries. In an interview, Raedler openly questioned whether AI can ultimately become a high-profit business to support the continued expansion of capital expenditure by tech giants in recent years. He suggested that investors shift their focus to "lagging" industries such as consumer staples and pharmaceuticals - industries whose basic business logic remains effective enough to avoid downside risks. This warning comes at a time when global AI concept stocks have experienced their most intense volatility since 2026. $725 billion bet: The "elephant in the room" of AI capital expenditure Data outlines the staggering scale of this bet. According to forecasts from multiple institutions, just Amazon, Alphabet, Microsoft, and Meta alone, the AI infrastructure capital expenditure in 2026 is expected to reach $725 billion, a 77% surge from 2025's $410 billion. Goldman Sachs further predicts that between 2025 and 2030, the cumulative capital expenditure in AI and data centers by super-scale cloud computing companies will reach as high as $53 trillion. However, the question of where the corresponding returns for massive investments lie is becoming a looming concern for all investors. Just this past Tuesday (June 23), the Nasdaq 100 index plummeted 3.3%, the Philadelphia Semiconductor Index crashed over 7%, with shares of Sandisk, Micron Technology plunging over 13%, ARM falling 10%, Qualcomm, and Western Digital dropping 8%. This sell-off spread from Asia-Pacific to US stocks - the South Korean KOSPI Index triggered a circuit breaker, plummeting nearly 10% in a single day, with Samsung Electronics and SK Hynix both plunging over 12%. "In the past three years, the AI industry has been running along a simple yet powerful logic line: the scarcer the computing power, the more reasonable the capital expenditure; the larger the capital expenditure, the higher the valuation; the higher the valuation, the easier the financing. This self-reinforcing cycle has been almost unquestioned," Raedler analyzed. "But as we enter mid-2026, every link in this logic chain is undergoing stress testing." "Who will foot the bill": Raedler's core questioning Raedler's argument points directly to the ultimate question of the AI industry chain: the payment willingness and pricing power of end-users. "The problem lies in 'who will foot the bill for all this capital expenditure driving Micron's demand growth?' The answer is, it will ultimately be borne by the end-users using AI," Raedler explained in the interview. "Where is their pricing power? How willing are they to pay? I think your current profit forecasts are too optimistic." He outlined a scenario that could trigger a chain reaction: "Once customers tell us they are not willing to pay as much for AI, once they switch their traffic to cheaper models, once the first super-scale data center operator says 'actually, I'm not sure I'm adding value here, let me lower the capital expenditure number' - then many defensive industries will experience explosive growth." Raedler believes that as AI service prices rise, their economic benefits are deteriorating, leading to consumer restrictions on usage; meanwhile, low-cost open-source models are disrupting the scarcity needed to maintain high-profit margins. Ultimately, profitability will depend on whether customers are willing to pay a premium for AI services or switch to lower-cost open-source alternatives. The confusion over the AI return on investment ratio has even spread to the ongoing Summer Davos Forum in Dalian. A forum titled "AI Everywhere, but Not Built in a Day" drew a full house. "Some companies have stated that in the past three years, AI has not had a measurable impact on their development," becoming a hot topic of discussion among attendees. PwC's latest report, "2026 AI Effectiveness Research Report," shows that only a few companies are leading in achieving financial benefits through AI. Jonas Prising, Chairman of Manpower GRC Group, summarized this dilemma of "high investment, low returns" as "phantom efficiency" - AI boosts individual efficiency but does not automatically organizational efficiency. Market shows signs of "defensive shift" Raedler's warning is not just an academic speculation. Market data indicates that funds have quietly shifted. In the chip stock sell-off on June 23, consumer staples and healthcare sectors significantly outperformed the market. Walmart, Procter & Gamble, and Johnson & Johnson all saw their stock prices rise by nearly 2%, while Merck and AstraZeneca also recorded gains. The State Street Consumer Staples ETF (XLP) rose by 1.8%. CITIC SEC US stock strategy pointed out that funds are rotating between super-scale cloud computing companies, semiconductor and software communication sectors, with defensive sectors such as healthcare and consumer receiving safe-haven inflows. This rotation is not an isolated incident. Wedbush analysts noted that the market turmoil at the end of February 2026 marked the end of the "AI honeymoon period," ushering in a more cautious investment environment. An analysis by Nasdaq also observed that most top AI stocks suddenly struggled, potentially shifting funds from high-risk growth stocks to more stable and resilient holdings. "The flight to safety reflects concerns in the market about the excessive enthusiasm for AI," said Chris Low of FHN Financial. Goldman Sachs echoes: The first giant to cut spending will trigger a comprehensive repricing Raedler is not the only voice on Wall Street issuing warnings. In a research report on June 23, Privorotsky, a strategist at Goldman Sachs' Global Banking and Markets division, warned that the AI market is like a stretched rubber band, and the market's persistent disregard for negative signals will eventually reach a tipping point. He pointed out a growing structural divergence: while super-scale cloud computing companies continue to ramp up spending commitments, their stock prices continue to lag behind the market; meanwhile, AI hardware stocks represented by NVIDIA and TSMC are defying the trend and rising. "Once any major tech giant takes the lead in cutting AI spending, the entire AI sector's valuation logic will face a comprehensive overhaul," Goldman Sachs warned. This assessment resonates with Raedler's arguments. Both strategists point to the same vulnerable point: when the gears of capital expenditure stop turning, the entire AI valuation system will lose its fulcrum. Disputing the pessimistic signals from shorts: Micron's financial report and Qualcomm's forecasts However, the bullish forces in the AI industry chain have not stayed silent. Just after the sharp drop on Tuesday, Micron Technology and Qualcomm released their performance outlook on Wednesday, leading to a market value increase of over $400 billion for chip stocks. Micron reported third-quarter revenue of $41.46 billion, far exceeding the expected $35.85 billion, while forecasting an adjusted earnings per share of $31 for the current quarter, compared to the market consensus of $25.84. The company stated that 16 strategic customers have signed long-term agreements totaling $22 billion, with remaining commitments totaling around $100 billion. Micron CEO Sanjay Mehrotra stated that AI-driven demand has caused the "most severe supply shortage in the past 15 years" as mentioned by Goldman Sachs, with a supply-demand gap of 4.9%, and the tight supply situation expected to continue beyond the 2027 calendar year. Qualcomm also announced at an investor day that it expects data center chip sales to reach $15 billion by 2029, officially entering the AI data center chip race. The optimistic forecasts from Micron and Qualcomm stand in stark contrast to Raedler's warning - on one side are semiconductor companies proving the existence of demand with orders and revenue data, and on the other side are strategists questioning the ability of end-users to foot the bill using economic logic.