AI Uncertainty Grows on Wall Street, Yet Investor Conviction Hits Record Highs

A record 82% of fund managers call AI the most crowded trade, yet conviction on both sides is at all-time highs as second-quarter earnings begin. Bank of America reports the Magnificent Seven have spent $234 billion in capex this year with free cash flow turning negative for the first time in 20 years, while AI-related compute capex now accounts for a record share of U.S. GDP. The polarization, fueled by trillion-dollar spending forecasts and conflicting data, could trigger volatile markets if earnings disappoint.

July 14, 2026, Inside AI — A paradox is gripping Wall Street as the second-quarter earnings season begins: uncertainty over artificial intelligence’s profit potential is deepening, yet both AI bulls and bears are digging in with unprecedented conviction. A record 82% of fund managers now call AI the most crowded trade, according to Bank of America’s latest survey, but roughly half still reject the bubble label. This polarization, fueled by trillion-dollar spending forecasts and conflicting data, could trigger a volatile second half. The bull case rests on a simple premise: massive AI capital expenditure will unlock productivity and profit gains that justify soaring stock prices. Bears counter that the buildout’s cost has ballooned beyond sustainable returns, with hyperscalers burning cash and turning to debt. Bank of America strategists illustrate this with a “generational transfer” of free cash flow from hyperscalers to chipmakers. The “Magnificent Seven” have spent $234 billion in capex this year, yet their stocks are flat as free cash flow turns negative for the first time in at least 20 years . “The vast sums hyperscalers are spending on AI infrastructure and capex are essentially flowing to semiconductor companies, which will enjoy an increasing share of future AI profits,” the strategists wrote. Meanwhile, bears highlight the soaring cost of compute. They argue demand for the priciest AI models will shift to cheaper, open-source alternatives, likely from China, eroding returns on the historic investment. The entrenchment is fueled by a lack of historical precedent. Unlike railroads or the internet, AI has no playbook, making it easier for each side to dismiss counterevidence. “Nobody knows how it will ultimately impact businesses, the workplace, jobs and the economy,” the report notes. But the stakes are rising. AI-related compute capex now accounts for a record share of U.S. GDP, and overall tech investment jumped 30% year-over-year, while all other capex fell. AI represents almost all net U.S. investment, according to Carlyle analysts. This concentration creates systemic risk. If the AI buildout is a zero-sum game, either hyperscalers’ cash flow must rebound or chipmakers’ growth must slow sharply. The SOX semiconductor index remains up 75% this year, but single-stock volatility has spiked for names like Intel , Qualcomm , and Oracle . Noah Weisberger, head of equities at BCA Research, said clients are increasingly worried about single-stock swings, adding that last month’s AI pullback showed investors had gotten “a little over their skis.” South Korea’s KOSPI , even more AI-concentrated than the S&P 500, offers a warning. Its three biggest drops since Lehman’s collapse—a 12% slump in March, 10% in June, and 9% on Monday driven by SK Hynix —all hit this year. As earnings season unfolds, both camps will scour results for validation. The deeper the conviction, the sharper the potential reckoning if the data disappoints.