The divide in market views over the artificial intelligence (AI) rally is widening. Big Tech's infrastructure investment is surging, and major semiconductor firms such as Samsung Electronics and Micron are reporting "record-breaking" earnings, yet investor sentiment remains clouded by doubts over the sustainability of the investment cycle. Even among global investment banks (IBs), debate is swirling over the direction of the AI rally and its leading stocks, drawing the market's attention to the "words" of major corporate chief executives set to report earnings through next week.
On Monday, U.S. political news outlet NOTUS reported that U.S. Treasury analysts, in a recent draft of a confidential report, warned that a failure to monetize the AI market could bring a shock to the broader financial system similar to the collapse of the dot-com bubble in the early 2000s. The analysis holds that the AI industry is more deeply connected to the broader U.S. economy than during the dot-com bubble. The report projected that if AI companies fail to demonstrate higher productivity and profitability than they do now, the shock will spread beyond the stock market to data center construction financing, the private credit market, cloud providers, semiconductor manufacturers, and the power and utilities industries.
Morgan Stanley, dubbed the "chip grim reaper," issued a similar warning. Morgan Stanley forecast that the global AI investment budgets set by major hyperscalers (large-scale cloud providers) would grow to $805 billion in 2026 and $1.116 trillion in 2027, while at the same time recommending reducing exposure to semiconductor stocks. This means the narrow, chip-concentrated stock rally has passed its peak, and that in the short term investors should increase exposure to hyperscalers. Morgan Stanley judged that semiconductors will ultimately remain subordinate to Big Tech's pace of investment, and that signals of a slowdown in investment pace can be detected in Meta's external sales of surplus AI computing capacity.
The U.S. Treasury and Morgan Stanley do not hold the position that the AI rally will meet a "catastrophe" immediately. The Treasury report analyzed that current major AI companies, unlike the speculative firms of the dot-com bubble era, have healthy finances and solid profitability in their core businesses. It suggests that even if the bubble deflates, they have a robust "basic fitness" to cushion the shock. Even so, concerns over a pullback in AI infrastructure investment—raised recently by Apple's product price hikes and Meta's entry into the "neocloud" business—show no signs of subsiding.
Not all of Wall Street has turned pessimistic. JPMorgan recently offered an optimistic view, calling the recent decline in semiconductor stock prices a "gift" from the market and a buying opportunity. It analyzes that global demand for AI chips remains solid while new production facilities will be difficult to secure until 2028. JPMorgan projected that as chronic supply shortages boost pricing power, the advantage of semiconductor manufacturers over hyperscalers will persist, and that global semiconductor stocks will set an all-time high during the second half of 2026. In fact, this year the five Big Tech firms' data center investment amounts to $848 billion, and semiconductor demand has grown more than fivefold since 2023.
Kim Sun-woo, an analyst at Meritz Securities, noted, "In the current cycle, a situation where memory supply falls woefully short of catching up with the pace of demand growth will persist at least through the fourth quarter of next year." Ryu Hyung-geun, an analyst at Daishin Securities, analyzed, "The average selling price (ASP) of high-bandwidth memory (HBM) is forecast to rise 91% year-on-year next year," adding, "Based on an absolute advantage, the profitability gap will narrow."
Experts agreed that since the strong second-quarter semiconductor earnings were already anticipated, "confidence" is needed to prove the sustainability of future AI infrastructure investment. This means that rather than earnings that have become "the past," what matters now are the specific contract details and future outlook that management will confirm through conference calls. Regarding the record-breaking earnings of Micron and Samsung Electronics, some in the market interpreted them as a typical sign of a late-stage cycle relying solely on price increases rather than volume growth.
As a result, observers expect that the direction of the AI market will be determined through major corporate earnings reports running until mid-July. On Monday in the U.S. market, Penguin Solutions, a data center infrastructure designer and key partner of Nvidia, released its earnings. Penguin's guidance and order backlog are indicators for confirming whether Big Tech firms are actually deploying funds for AI infrastructure construction. On the 10th, TSMC's June revenue announcement and SK hynix's listing of American depositary receipts (ADRs) in the U.S. are scheduled. On the 15th, ASML reports earnings. On the 16th, the official quarterly earnings reports of TSMC and storage device maker Seagate are scheduled.
Seo Sang-young, managing director at Mirae Asset Securities, said, "If Penguin Solutions' earnings show that data center construction is expanding, semiconductor jitters could ease," adding, "As for TSMC's earnings, a short-term increase in volatility is inevitable, but the possibility of the market contracting significantly is limited."
