In a note to clients reviewed by Fortune, BofA strategists declared that “doubts around the AI revolution are emerging,” with the market narrative rapidly shifting from an “upside-only” perspective to serious concerns that AI is a “double-edged sword.” Chief among these new fears is the growing realization that AI might not universally boost corporate profits—it might actively destroy them.

BofA highlighted several large “downside risks” that are, frankly, bumming out the AI trade. Traders are confronting a world of cannibalization, capital expenditure cuts, and other various AI-related monsters.

The threat of profit cannibalization

BofA points out a glaring blind spot in current market expectations. The sell side consensus currently projects a staggering 17% compound annualized earnings per share (EPS) growth for global equities over the next five years. This lies at the heart of what strategists call the “cannibalization” paradox.

Corporate margins are already at all-time highs, BofA strategists noted, complicating this 17% forecast. Historically, double-digit compound EPS growth has only been achieved when margins were depressed. These “optimistic” expectations therefore assume that tech EPS growth from AI adoption will be sustainable and at the same time that this growth will not “cannibalize existing profit pools.” That is exactly where investors are showing “the largest rethink,” with sharp declines in AI-vulnerable sectors such as software, insurance, and wealth management, and a flight to safer sectors with reliance on physical assets (e.g., mining, utilities, and chemicals).

This painful reality is already hitting the market. For instance, SAP, an early software champion that traded as an AI beneficiary, recently saw its massive rally abruptly reversed. The selloff in software-as-a-service (SaaS) stocks, reaching more than $2 trillion in over two weeks and change, was dubbed the “SaaSpocalypse” as a result.

Labor market fragility and capex cracks

Beyond direct business cannibalization, BofA warned that the AI rollout poses severe macroeconomic risks. The U.S. labor market is already flashing warning signs, with three-month payroll growth hovering at an anemic 0.1%—a level of zero job growth historically associated with the end of past equity bull runs. This includes both 2000 and 2007, when bubbles bursting coincided with painful recessions.

BofA cautions that AI-related productivity gains could further depress corporate demand for labor, worsening this weakness. Ironically, massive job losses could backfire on the tech giants themselves, as the hyperscalers funding the AI boom rely heavily on consumer advertising revenues, which demand a healthy, employed consumer base. The implicit assumption from consensus market and earnings projections, analysts argue, is that weakness in the labor market is driven by temporary headwinds including the trade picture, the immigration crackdown, and over-hiring during the Great Resignation post-pandemic era, but what if those assumptions are wrong?

Furthermore, the debt-fueled AI investment spree is showing distinct signs of strain. For the first time in the history of BofA’s Global Fund Manager Survey, investors reported that they believe companies are overinvesting and should cut back on capital expenditures (capex). Corporate bond spreads for U.S. hyperscalers have hit a three-year high, and private credit platforms deeply exposed to software debt are buckling under the pressure, highlighted by Blue Owl suspending retail investor redemptions for one of its funds this week. BofA’s U.S. equity strategy team expects these headwinds to culminate in an “AI air pocket” later in 2026.

Finding ‘AI hiding places’

Adding to the gloom, BofA warns that the promised economic boom from AI may fail to materialize. While equity markets are currently pricing in near 3% U.S. productivity growth, forecasters such as the Congressional Budget Office project that AI will only boost productivity by a paltry 0.1% annually over the next decade. On this note, fresh data on Friday revealed the economy grew much less than expected in the fourth quarter of 2025, just 1.4%, after surprising economists with stronger growth for much of the rest of the year.

In response to this shifting landscape, BofA cautioned that AI infrastructure providers—such as semiconductors, capital goods, and construction materials—now look dangerously “stretched,” trading at all-time-high relative prices and peak earnings expectations. Declaring them highly vulnerable to AI capex disappointments, BofA has aggressively downgraded the semiconductor sector to an underweight rating.

In the hyperscaler space, cracks are visible if you squint amid rumors that OpenAI is planning an IPO in late 2026 as funding shortfalls emerge for its voracious hunger for compute. Nvidia CEO Jensen Huang said in early February that his agreement to invest $100 billion in OpenAI was “never a commitment” while insisting that he was still fully enjoying his partnership with Sam Altman’s company. Owen Lamont, a portfolio manager at Acadian Asset Management and a former University of Chicago finance professor, told Fortune earlier this month that the current picture in markets qualifies as a bubble by three of his four bespoke metrics, with the only missing element being a flood of IPOs.

Instead of chasing the AI rally, BofA advises investors to seek shelter in “AI hiding places.” The firm is maintaining overweight positions in defensive sectors with limited AI disruption risk, such as consumer staples, telecoms, and chemicals. Overall, the bank remains decidedly negative on European equities, projecting a 15% downside by the second quarter as the painful reality of a cannibalized, double-edged AI market fully sets in.